{ "openapi": "3.0.0", "info": { "title": "MVSEP API", "version": "1.0.0", "description": "# MVSEP API Documentation\n\n## Overview\n\nThe MVSEP (Music Voice Separation) API provides a comprehensive platform for audio source separation, allowing you to separate vocals, instruments, and other audio components from mixed audio files. This API enables developers to integrate professional-grade audio separation capabilities into their applications.\n\n**Base URL:** `https://mvsep.com/api`\n\n## Key Features\n\n- **Audio Separation**: Create and retrieve audio separation tasks with various algorithms\n \n- **User Management**: Register, login, and manage user profiles\n \n- **Premium Features**: Enable/disable premium functionality for enhanced processing\n \n- **Separation History**: Track and retrieve past separation jobs\n \n- **Quality Control**: Add and manage quality checker entries\n \n- **Algorithm Discovery**: Browse available separation algorithms\n \n- **News & Updates**: Access platform news and announcements\n \n\n## Getting Started\n\n1. **Register an Account**: Use the `/app/register` endpoint to create a new account\n \n2. **Login**: Authenticate using the `/app/login` endpoint to receive your API token\n \n3. **Set API Token**: Store the `api_token` in your environment variables for authenticated requests\n \n4. **Explore Algorithms**: Check available separation algorithms via `/app/algorithms`\n \n5. **Create Separation**: Submit audio files for processing using `/separation/create`\n \n6. **Retrieve Results**: Poll the separation status and download results with `/separation/get`\n \n\n## Authentication\n\nMost endpoints require authentication using an API token. The token should be passed as a query parameter:\n\n```\n?api_token={{api_token}}\n\n ```\n\n**Environment Variable:**\n\n- `api_token` - Your personal API authentication token (obtained after login)\n \n\n### How to Authenticate\n\n1. Register or login to receive your API token\n \n2. Store the token in the `api_token` environment variable\n \n3. The token will be automatically included in authenticated requests\n \n\n**Example:**\n\n```\nGET https://mvsep.com/api/app/user?api_token={{api_token}}\n\n ```\n\n## API Endpoints\n\n### Authentication & User Management\n\n#### Register\n\n- **Endpoint**: `POST /app/register`\n \n- **Description**: Create a new user account\n \n- **Authentication**: Not required\n \n\n#### Login\n\n- **Endpoint**: `POST /app/login`\n \n- **Description**: Authenticate and receive API token\n \n- **Authentication**: Not required\n \n- **Returns**: API token for subsequent requests\n \n\n#### User Profile\n\n- **Endpoint**: `GET /app/user`\n \n- **Description**: Retrieve current user information\n \n- **Authentication**: Required (`api_token`)\n \n\n#### Demo\n\n- **Endpoint**: `GET /app/demo`\n \n- **Description**: Access demo functionality\n \n- **Authentication**: Not required\n \n\n---\n\n### Audio Separation\n\n#### Get Available Algorithms\n\n- **Endpoint**: `GET /app/algorithms`\n \n- **Description**: List all available audio separation algorithms\n \n- **Authentication**: Not required\n \n- **Use Case**: Discover which separation models are available before creating a separation task\n \n\n#### Create Separation\n\n- **Endpoint**: `POST /separation/create`\n \n- **Description**: Submit an audio file for separation processing\n \n- **Authentication**: Required\n \n- **Process**: Upload audio file and specify desired separation algorithm\n \n\n#### Get Separation Results\n\n- **Endpoint**: `GET /separation/get`\n \n- **Description**: Retrieve separation results by hash\n \n- **Authentication**: Not required\n \n- **Parameters**:\n \n - `hash` - Unique identifier for the separation task\n \n- **Example**: `/separation/get?hash=20240119124024-413c481b80-10-luke-s-theme-a-madman-s-symphony.mp3`\n \n\n#### Separation History\n\n- **Endpoint**: `GET /app/separation_history`\n \n- **Description**: Retrieve user's past separation jobs\n \n- **Authentication**: Required (`api_token`)\n \n- **Parameters**:\n \n - `start` - Starting index for pagination (e.g., 1)\n \n - `limit` - Number of results to return (e.g., 2)\n \n- **Example**: `/app/separation_history?api_token={{api_token}}&start=1&limit=2`\n \n\n---\n\n### Profile Management\n\nThe Profile folder contains endpoints for managing user preferences and features:\n\n#### Enable Premium\n\n- **Endpoint**: `POST /app/enable_premium`\n \n- **Description**: Activate premium features for the user account\n \n- **Authentication**: Required\n \n\n#### Disable Premium\n\n- **Endpoint**: `POST /app/disable_premium`\n \n- **Description**: Deactivate premium features\n \n- **Authentication**: Required\n \n\n#### Enable Long Filenames\n\n- **Endpoint**: `POST /app/enable_long_filenames`\n \n- **Description**: Enable support for long filename processing\n \n- **Authentication**: Required\n \n\n#### Disable Long Filenames\n\n- **Endpoint**: `POST /app/disable_long_filenames`\n \n- **Description**: Disable long filename support\n \n- **Authentication**: Required\n \n\n---\n\n### Quality Checker\n\nQuality control endpoints for managing separation quality entries:\n\n#### Add Entry\n\n- **Endpoint**: `POST /quality_checker/add`\n \n- **Description**: Add a new quality checker entry\n \n- **Authentication**: Required\n \n\n#### Delete Entry\n\n- **Endpoint**: `POST /quality_checker/delete`\n \n- **Description**: Remove a quality checker entry\n \n- **Authentication**: Required\n \n \n\n---\n\n### News & Information\n\n#### Get News\n\n- **Endpoint**: `GET /app/news`\n \n- **Description**: Retrieve platform news and announcements\n \n- **Authentication**: Not required\n \n- **Parameters**:\n \n - `lang` - Language code (e.g., 'ru', 'en')\n \n - `start` - Starting index for pagination\n \n - `limit` - Number of news items to return\n \n- **Example**: `/app/news?lang=ru&start=2&limit=1`\n \n\n---\n\n## Typical Workflow\n\n1. **Register/Login** → Obtain `api_token`\n \n2. **Get Algorithms** → Choose separation algorithm\n \n3. **Create Separation** → Upload audio file\n \n4. **Poll Status** → Check separation progress\n \n5. **Get Results** → Download separated audio tracks\n \n6. **View History** → Access past separations\n \n\n## Rate Limits & Pricing\n\nPremium accounts may have different rate limits and access to advanced algorithms. Use the premium enable/disable endpoints to manage subscription features.\n\n## Support & Resources\n\nFor additional information about audio separation algorithms, processing times, and best practices, please refer to the MVSEP platform documentation at [https://mvsep.com](https://mvsep.com).\n\n## Environment Variables\n\nMake sure to configure the following environment variable in your Postman environment:\n\n| Variable | Description | Example |\n| --- | --- | --- |\n| `api_token` | Your API authentication token | `your-token-here` |\n\n---\n\n**API Version**: 1.0 \n**Last Updated**: 2024 \n**Contact**: For support, visit [https://mvsep.com](https://mvsep.com)" }, "servers": [ { "url": "https://mvsep.com" } ], "paths": { "/api/app/enable_long_filenames": { "post": { "summary": "Enable Long Filenames", "responses": { "200": { "description": "Enable Long Filenames", "headers": { "Date": { "schema": { "type": "string" }, "example": "Sat, 30 Mar 2024 06:31:36 GMT" }, "Server": { "schema": { "type": "string" }, "example": "Apache/2.4.41 (Ubuntu)" }, "Cache-Control": { "schema": { "type": "string" }, "example": "no-cache, private" }, "X-RateLimit-Limit": { "schema": { "type": "integer" }, "example": "120" }, "X-RateLimit-Remaining": { "schema": { "type": "integer" }, "example": "119" }, "Keep-Alive": { "schema": { "type": "string" }, "example": "timeout=5, max=100" }, "Connection": { "schema": { "type": "string" }, "example": "Keep-Alive" }, "Transfer-Encoding": { "schema": { "type": "string" }, "example": "chunked" } }, "content": { "application/json": { "schema": { "type": "object", "properties": { "success": { "type": "boolean" }, "message": { "type": "string" } } }, "example": { "success": true, "message": "successfully enabled" } } } } }, "tags": [ "Profile" ] } }, "/api/app/enable_premium": { "post": { "summary": "Enable Premium", "responses": { "200": { "description": "Enable Premium", "headers": { "Date": { "schema": { "type": "string" }, "example": "Sat, 30 Mar 2024 06:33:22 GMT" }, "Server": { "schema": { "type": "string" }, "example": "Apache/2.4.41 (Ubuntu)" }, "Cache-Control": { "schema": { "type": "string" }, "example": "no-cache, private" }, "X-RateLimit-Limit": { "schema": { "type": "integer" }, "example": "120" }, "X-RateLimit-Remaining": { "schema": { "type": "integer" }, "example": "118" }, "Keep-Alive": { "schema": { "type": "string" }, "example": "timeout=5, max=100" }, "Connection": { "schema": { "type": "string" }, "example": "Keep-Alive" }, "Transfer-Encoding": { "schema": { "type": "string" }, "example": "chunked" } }, "content": { "application/json": { "schema": { "type": "object", "properties": { "success": { "type": "boolean" }, "message": { "type": "string" } } }, "example": { "success": true, "message": "successfully enabled" } } } } }, "tags": [ "Profile" ] } }, "/api/app/disable_premium": { "post": { "summary": "Disable Premium", "responses": { "200": { "description": "Disable Premium", "headers": { "Date": { "schema": { "type": "string" }, "example": "Sat, 30 Mar 2024 06:44:04 GMT" }, "Server": { "schema": { "type": "string" }, "example": "Apache/2.4.41 (Ubuntu)" }, "Cache-Control": { "schema": { "type": "string" }, "example": "no-cache, private" }, "X-RateLimit-Limit": { "schema": { "type": "integer" }, "example": "120" }, "X-RateLimit-Remaining": { "schema": { "type": "integer" }, "example": "119" }, "Keep-Alive": { "schema": { "type": "string" }, "example": "timeout=5, max=100" }, "Connection": { "schema": { "type": "string" }, "example": "Keep-Alive" }, "Transfer-Encoding": { "schema": { "type": "string" }, "example": "chunked" } }, "content": { "application/json": { "schema": { "type": "object", "properties": { "success": { "type": "boolean" }, "message": { "type": "string" } } }, "example": { "success": true, "message": "successfully disabled" } } } } }, "tags": [ "Profile" ] } }, "/api/app/user": { "get": { "summary": "User", "parameters": [ { "name": "api_token", "in": "query", "schema": { "type": "string" }, "example": "{{api_token}}" } ], "responses": {}, "tags": [ "Profile" ] } }, "/api/app/disable_long_filenames": { "post": { "summary": "Disable Long FIlenames", "responses": { "200": { "description": "Disable Long Filenames", "headers": { "Date": { "schema": { "type": "string" }, "example": "Sat, 30 Mar 2024 06:43:00 GMT" }, "Server": { "schema": { "type": "string" }, "example": "Apache/2.4.41 (Ubuntu)" }, "Cache-Control": { "schema": { "type": "string" }, "example": "no-cache, private" }, "X-RateLimit-Limit": { "schema": { "type": "integer" }, "example": "120" }, "X-RateLimit-Remaining": { "schema": { "type": "integer" }, "example": "119" }, "Keep-Alive": { "schema": { "type": "string" }, "example": "timeout=5, max=100" }, "Connection": { "schema": { "type": "string" }, "example": "Keep-Alive" }, "Transfer-Encoding": { "schema": { "type": "string" }, "example": "chunked" } }, "content": { "application/json": { "schema": { "type": "object", "properties": { "success": { "type": "boolean" }, "message": { "type": "string" } } }, "example": { "success": true, "message": "successfully disabled" } } } } }, "tags": [ "Profile" ] } }, "/api/app/separation_history": { "get": { "summary": "Separation History", "parameters": [ { "name": "api_token", "in": "query", "schema": { "type": "string" }, "example": "{{api_token}}" }, { "name": "start", "in": "query", "schema": { "type": "integer" }, "example": "1" }, { "name": "limit", "in": "query", "schema": { "type": "integer" }, "example": "2" } ], "responses": { "200": { "description": "Separation History", "headers": { "Date": { "schema": { "type": "string" }, "example": "Fri, 10 May 2024 08:02:47 GMT" }, "Server": { "schema": { "type": "string" }, "example": "Apache/2.4.52 (Ubuntu)" }, "Cache-Control": { "schema": { "type": "string" }, "example": "no-cache, private" }, "X-RateLimit-Limit": { "schema": { "type": "integer" }, "example": "120" }, "X-RateLimit-Remaining": { "schema": { "type": "integer" }, "example": "119" }, "Keep-Alive": { "schema": { "type": "string" }, "example": "timeout=5, max=100" }, "Connection": { "schema": { "type": "string" }, "example": "Keep-Alive" }, "Transfer-Encoding": { "schema": { "type": "string" }, "example": "chunked" } }, "content": { "application/json": { "schema": { "type": "object", "properties": { "success": { "type": "boolean" }, "data": { "type": "array", "items": { "type": "object", "properties": { "id": { "type": "integer" }, "hash": { "type": "string" }, "created_at": { "type": "string", "format": "style" }, "updated_at": { "type": "string" }, "job_exists": { "type": "boolean" }, "credits": { "type": "integer" }, "time_left": { "type": "integer" }, "algorithm": { "type": "object", "properties": { "name": { "type": "string" }, "algorithm_group_id": { "type": "integer" }, "orientation": { "type": "integer" }, "render_id": { "type": "integer" }, "order_id": { "type": "integer" }, "created_at": { "type": "string", "format": "style" }, "updated_at": { "type": "string" }, "algorithm_group": { "type": "object", "properties": { "id": { "type": "integer" }, "name": { "type": "string" }, "created_at": { "type": "string", "format": "style" }, "updated_at": { "type": "string" } } } } } } } } } }, "example": { "success": true, "data": [ { "id": 1074359, "hash": "20240212161927-413c481b80-10-luke-s-theme-a-madman-s-symphony.mp3", "created_at": "2024-02-12 16:19:29", "updated_at": "2024-02-12 16:19:29", "job_exists": true, "credits": 2, "time_left": -2031, "algorithm": { "name": "Demucs4 HT (vocals, drums, bass, other)", "algorithm_group_id": 2, "orientation": 1, "render_id": 20, "order_id": 1, "created_at": "2022-11-28 09:37:27", "updated_at": "2023-03-13 02:52:03", "algorithm_group": { "id": 2, "name": "HQ Models", "created_at": "2022-11-28 05:54:31", "updated_at": "2022-11-28 05:54:31" } } }, { "id": 1074358, "hash": "20240212095132-413c481b80-10-luke-s-theme-a-madman-s-symphony.mp3", "created_at": "2024-02-12 09:51:34", "updated_at": "2024-02-12 09:51:34", "job_exists": true, "credits": 2, "time_left": -2038, "algorithm": { "name": "Demucs4 HT (vocals, drums, bass, other)", "algorithm_group_id": 2, "orientation": 1, "render_id": 20, "order_id": 1, "created_at": "2022-11-28 09:37:27", "updated_at": "2023-03-13 02:52:03", "algorithm_group": { "id": 2, "name": "HQ Models", "created_at": "2022-11-28 05:54:31", "updated_at": "2022-11-28 05:54:31" } } }, { "id": 1074357, "hash": "20240212075543-a274f83da2-4-un-poco-de-tu-amor-5.flac", "created_at": "2024-02-12 07:55:44", "updated_at": "2024-02-12 07:55:44", "job_exists": true, "credits": 3, "time_left": -2040, "algorithm": { "name": "Demucs4 HT (vocals, drums, bass, other)", "algorithm_group_id": 2, "orientation": 1, "render_id": 20, "order_id": 1, "created_at": "2022-11-28 09:37:27", "updated_at": "2023-03-13 02:52:03", "algorithm_group": { "id": 2, "name": "HQ Models", "created_at": "2022-11-28 05:54:31", "updated_at": "2022-11-28 05:54:31" } } }, { "id": 1074356, "hash": "20240206094315-fb58a4ed91-haliene-dream-in-color-0.mp3", "created_at": "2024-02-06 09:43:15", "updated_at": "2024-02-06 09:43:15", "job_exists": true, "credits": 3, "time_left": -2182, "algorithm": { "name": "Demucs4 HT (vocals, drums, bass, other)", "algorithm_group_id": 2, "orientation": 1, "render_id": 20, "order_id": 1, "created_at": "2022-11-28 09:37:27", "updated_at": "2023-03-13 02:52:03", "algorithm_group": { "id": 2, "name": "HQ Models", "created_at": "2022-11-28 05:54:31", "updated_at": "2022-11-28 05:54:31" } } }, { "id": 1074355, "hash": "20240206094008-b2e84deb3a-haliene-dream-in-color.mp3", "created_at": "2024-02-06 09:40:08", "updated_at": "2024-02-06 09:40:08", "job_exists": true, "credits": 3, "time_left": -2182, "algorithm": { "name": "Demucs4 HT (vocals, drums, bass, other)", "algorithm_group_id": 2, "orientation": 1, "render_id": 20, "order_id": 1, "created_at": "2022-11-28 09:37:27", "updated_at": "2023-03-13 02:52:03", "algorithm_group": { "id": 2, "name": "HQ Models", "created_at": "2022-11-28 05:54:31", "updated_at": "2022-11-28 05:54:31" } } }, { "id": 1074354, "hash": "20240205145513-c46e6a674c-skylar-grey-im-coming-home-ano-remix-z2-fm.mp3", "created_at": "2024-02-05 14:55:13", "updated_at": "2024-02-05 14:55:13", "job_exists": true, "credits": 3, "time_left": -2201, "algorithm": { "name": "MDX23C (vocals, instrumental)", "algorithm_group_id": 2, "orientation": 0, "render_id": 25, "order_id": 2, "created_at": "2023-07-05 15:32:54", "updated_at": "2023-07-05 15:33:54", "algorithm_group": { "id": 2, "name": "HQ Models", "created_at": "2022-11-28 05:54:31", "updated_at": "2022-11-28 05:54:31" } } }, { "id": 1074351, "hash": "20240205144711-c46e6a674c-skylar-grey-im-coming-home-ano-remix-z2-fm.mp3", "created_at": "2024-02-05 14:47:11", "updated_at": "2024-02-05 14:47:11", "job_exists": true, "credits": 3, "time_left": -2201, "algorithm": { "name": "MDX23C (vocals, instrumental)", "algorithm_group_id": 2, "orientation": 0, "render_id": 25, "order_id": 2, "created_at": "2023-07-05 15:32:54", "updated_at": "2023-07-05 15:33:54", "algorithm_group": { "id": 2, "name": "HQ Models", "created_at": "2022-11-28 05:54:31", "updated_at": "2022-11-28 05:54:31" } } }, { "id": 1074350, "hash": "20240124130755-f40709507a-sample-15s-4.mp3", "created_at": "2024-01-24 13:07:55", "updated_at": "2024-01-24 13:07:55", "job_exists": true, "credits": 1, "time_left": -2490, "algorithm": { "name": "Demucs4 HT (vocals, drums, bass, other)", "algorithm_group_id": 2, "orientation": 1, "render_id": 20, "order_id": 1, "created_at": "2022-11-28 09:37:27", "updated_at": "2023-03-13 02:52:03", "algorithm_group": { "id": 2, "name": "HQ Models", "created_at": "2022-11-28 05:54:31", "updated_at": "2022-11-28 05:54:31" } } }, { "id": 1074349, "hash": "20240124130325-514b0abb0f-sample-3s-26.mp3", "created_at": "2024-01-24 13:03:25", "updated_at": "2024-01-24 13:03:25", "job_exists": true, "credits": 1, "time_left": -2490, "algorithm": { "name": "Demucs4 HT (vocals, drums, bass, other)", "algorithm_group_id": 2, "orientation": 1, "render_id": 20, "order_id": 1, "created_at": "2022-11-28 09:37:27", "updated_at": "2023-03-13 02:52:03", "algorithm_group": { "id": 2, "name": "HQ Models", "created_at": "2022-11-28 05:54:31", "updated_at": "2022-11-28 05:54:31" } } }, { "id": 1074348, "hash": "20240119124024-413c481b80-10-luke-s-theme-a-madman-s-symphony.mp3", "created_at": "2024-01-19 12:40:25", "updated_at": "2024-01-19 15:49:37", "job_exists": true, "credits": 9, "time_left": -2611, "algorithm": { "name": "Ensemble 8 models (vocals, bass, drums, other)", "algorithm_group_id": 3, "orientation": 2, "render_id": 28, "order_id": 2, "created_at": "2023-03-17 11:50:21", "updated_at": "2023-08-07 23:36:52", "algorithm_group": { "id": 3, "name": "Ensemble", "created_at": "2022-11-28 05:54:38", "updated_at": "2022-11-28 05:54:38" } } } ] } } } } }, "tags": [ "Profile" ] } }, "/api/quality_checker/add": { "post": { "summary": "Add Entry", "responses": { "200": { "description": "Add Entry", "headers": { "Server": { "schema": { "type": "string" }, "example": "nginx" }, "Date": { "schema": { "type": "string" }, "example": "Wed, 22 Jan 2025 11:27:18 GMT" }, "Transfer-Encoding": { "schema": { "type": "string" }, "example": "chunked" }, "Connection": { "schema": { "type": "string" }, "example": "keep-alive" }, "Vary": { "schema": { "type": "string" }, "example": "Accept-Encoding" }, "Cache-Control": { "schema": { "type": "string" }, "example": "private, must-revalidate" }, "pragma": { "schema": { "type": "string" }, "example": "no-cache" }, "expires": { "schema": { "type": "integer" }, "example": "-1" }, "X-RateLimit-Limit": { "schema": { "type": "integer" }, "example": "120" }, "X-RateLimit-Remaining": { "schema": { "type": "integer" }, "example": "119" }, "Content-Encoding": { "schema": { "type": "string" }, "example": "gzip" } }, "content": { "application/json": { "schema": { "type": "object", "properties": { "success": { "type": "boolean" }, "data": { "type": "object", "properties": { "id": { "type": "integer" }, "link": { "type": "string", "format": "uri" } } } } }, "example": { "success": true, "data": { "id": 7772, "link": "https://mvsep.com/api/quality_checker/entry?id=7772" } } } } } }, "tags": [ "Quality Checker" ] } }, "/api/quality_checker/delete": { "post": { "summary": "Delete Entry", "responses": { "200": { "description": "Delete Entry", "headers": { "Date": { "schema": { "type": "string" }, "example": "Wed, 22 Jan 2025 10:44:53 GMT" }, "Server": { "schema": { "type": "string" }, "example": "Apache/2.4.52 (Ubuntu)" }, "Cache-Control": { "schema": { "type": "string" }, "example": "no-cache, private" }, "X-RateLimit-Limit": { "schema": { "type": "integer" }, "example": "120" }, "X-RateLimit-Remaining": { "schema": { "type": "integer" }, "example": "119" }, "Keep-Alive": { "schema": { "type": "string" }, "example": "timeout=5, max=100" }, "Connection": { "schema": { "type": "string" }, "example": "Keep-Alive" }, "Transfer-Encoding": { "schema": { "type": "string" }, "example": "chunked" } }, "content": { "application/json": { "schema": { "type": "object", "properties": { "success": { "type": "boolean" }, "data": { "type": "object", "properties": { "message": { "type": "string" } } } } }, "example": { "success": true, "data": { "message": "successfully deleted" } } } } } }, "tags": [ "Quality Checker" ] } }, "/api/app/algorithms": { "get": { "summary": "Algorithms", "responses": { "200": { "description": "Algorithms", "headers": { "Date": { "schema": { "type": "string" }, "example": "Mon, 05 Feb 2024 14:31:08 GMT" }, "Server": { "schema": { "type": "string" }, "example": "Apache/2.4.52 (Ubuntu)" }, "Cache-Control": { "schema": { "type": "string" }, "example": "no-cache, private" }, "X-RateLimit-Limit": { "schema": { "type": "integer" }, "example": "60" }, "X-RateLimit-Remaining": { "schema": { "type": "integer" }, "example": "59" }, "Keep-Alive": { "schema": { "type": "string" }, "example": "timeout=5, max=100" }, "Connection": { "schema": { "type": "string" }, "example": "Keep-Alive" }, "Transfer-Encoding": { "schema": { "type": "string" }, "example": "chunked" } }, "content": { "application/json": { "schema": { "type": "array", "items": { "type": "object", "properties": { "id": { "type": "integer" }, "name": { "type": "string" }, "algorithm_group_id": { "type": "integer" }, "orientation": { "type": "integer" }, "render_id": { "type": "integer" }, "order_id": { "type": "integer" }, "created_at": { "type": "string", "format": "style" }, "updated_at": { "type": "string" }, "additional_fields": { "nullable": true }, "price_coefficient": { "type": "integer" }, "is_active": { "type": "integer" }, "algorithm_fields": { "type": "array", "items": { "type": "object", "properties": { "id": { "type": "integer" }, "name": { "type": "string" }, "text": { "type": "string" }, "algorithm_id": { "type": "integer" }, "options": { "type": "string", "format": "style" }, "created_at": { "type": "string" }, "updated_at": { "type": "string", "format": "style" }, "default_key": { "type": "string", "format": "utc-millisec" }, "server_key": { "type": "string" } } } }, "algorithm_descriptions": { "type": "array", "items": { "type": "object", "properties": { "id": { "type": "integer" }, "short_description": { "type": "string", "format": "style" }, "long_description": { "type": "string", "format": "style" }, "lang": { "type": "string" }, "algorithm_id": { "type": "integer" }, "created_at": { "type": "string", "format": "style" }, "updated_at": { "type": "string" } } } } } } }, "example": [ { "id": 27, "name": "Ensemble 4 models (vocals, instrum)", "algorithm_group_id": 3, "orientation": 2, "render_id": 26, "order_id": 1, "created_at": "2023-07-05 16:18:13", "updated_at": "2024-01-19 12:36:05", "additional_fields": null, "price_coefficient": 2, "is_active": 0, "algorithm_fields": [ { "id": 18, "name": "add_opt1", "text": "Output files", "algorithm_id": 27, "options": "{\n \"0\": \"Standard set\",\n \"1\": \"Include intermediate results and max_fft, min_fft\"\n}", "created_at": "2024-01-03 14:57:46", "updated_at": "2024-01-03 14:57:46", "default_key": "0", "server_key": "detailed" } ], "algorithm_descriptions": [ { "id": 9, "short_description": "Ансамбль 4 моделей: UVR-MDX-NET-Voc_FT, Demucs4 Vocals 2023, лучшую модель MDX23C и VitLarge23. Алгоритм даёт максимально возможное качество для вокальной и инструментальной дорожек.", "long_description": "

Ансамбль 4 моделей: UVR-MDX-NET-Voc_FT, Demucs4 Vocals 2023, лучшую модель MDX23C и VitLarge23. Алгоритм даёт максимально возможное качество для вокальной и инструментальной дорожек.

\r\n

Таблица качества

\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n
Algorithm nameMultisong datasetSynth datasetMDX23 Leaderboard
SDR VocalsSDR InstrumentalSDR VocalsSDR InstrumentalSDR Vocals
MDX23C Ensemble10.4416.7412.7612.4611.17
", "lang": "ru", "algorithm_id": 27, "created_at": "2023-07-05 23:12:31", "updated_at": "2023-11-20 16:57:32" }, { "id": 10, "short_description": "Ensemble of UVR-MDX-NET-Voc_FT, Demucs4 Vocals 2023, best MDX23C model and VitLarge23 model. Algorithm gives the highest possible quality for vocal and instrumental stems.", "long_description": "

Ensemble of UVR-MDX-NET-Voc_FT, Demucs4 Vocals 2023, best MDX23C model and VitLarge23 model. Algorithm gives the highest possible quality for vocal and instrumental stems.

\r\n

Quality table

\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n
Algorithm nameMultisong datasetSynth datasetMDX23 Leaderboard
SDR VocalsSDR InstrumentalSDR VocalsSDR InstrumentalSDR Vocals
Ensemble of 4 models10.4416.7412.7612.4611.17
", "lang": "en", "algorithm_id": 27, "created_at": "2023-07-05 23:13:11", "updated_at": "2023-11-20 16:57:06" } ] }, { "id": 22, "name": "Ensemble 8 models (vocals, bass, drums, other)", "algorithm_group_id": 3, "orientation": 2, "render_id": 28, "order_id": 2, "created_at": "2023-03-17 11:50:21", "updated_at": "2023-08-07 23:36:52", "additional_fields": null, "price_coefficient": 4, "is_active": 1, "algorithm_fields": [ { "id": 19, "name": "add_opt1", "text": "Output files", "algorithm_id": 22, "options": "{\n \"0\": \"Standard set\",\n \"1\": \"Include intermediate results and max_fft, min_fft\"\n}", "created_at": "2024-01-03 15:28:01", "updated_at": "2024-01-03 15:28:01", "default_key": "0", "server_key": "detailed" } ], "algorithm_descriptions": [ { "id": 6, "short_description": "This ensemble is based on algorithm which took 2nd place at Music Demixing Track of Sound Demixing Challenge 2023. The main changes comparing to contest version is much better vocal models, which is used here. UVR-MDX-NET-Voc_FT, Demucs4 Vocals 2023, best MDX23C model and VitLarge23 model. For stems 'bass', 'drums' and 'other' we us the following 4 models: demucsht_ft, deumcs_ht, demucs_6s and demucs_mmi.", "long_description": "

This ensemble is based on algorithm which took 2nd place at Music Demixing Track of Sound Demixing Challenge 2023. The main changes comparing to contest version is much better vocal models, which is used here. We use following 4 models for vocals: UVR-MDX-NET-Voc_FT, Demucs4 Vocals 2023, best MDX23C model and VitLarge23 model. For stems 'bass', 'drums' and 'other' we us the following 4 models: demucsht_ft, deumcs_ht, demucs_6s and demucs_mmi. Initial winning model available here: https://github.com/ZFTurbo/MVSEP-MDX23-music-separation-model

\r\n

Quality table

\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n
Multisong datasetSynth dataset
SDR BassSDR DrumsSDR OtherSDR VocalsSDR InstrumentalSDR VocalsSDR Instrumental
12.5211.737.0110.3016.6012.6712.38
\r\n

 

\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n
MDX23 Leaderboard
SDR BassSDR DrumsSDR OtherSDR Vocals
9.9379.5597.28011.093
", "lang": "en", "algorithm_id": 22, "created_at": "2023-06-12 11:53:43", "updated_at": "2023-10-04 03:07:24" }, { "id": 13, "short_description": "Ансамбль основан на алгоритме, занявшем 2-е место на Music Demixing Track of Sound Demixing Challenge 2023. Основное отличие от конкурсной версии — гораздо более совершенные вокальные модели, которые здесь используются. Для вокала мы используем следующие 4 модели: UVR-MDX-NET-Voc_FT, Demucs4 Vocals 2023, лучшую модель MDX23C и VitLarge23. Для дорожек «bass», «drums» и «other» мы используем следующие 4 модели: demucsht_ft, demucs_ht, demucs_6s и demucs_mmi.", "long_description": "

Ансамбль основан на алгоритме, занявшем 2-е место на Music Demixing Track of Sound Demixing Challenge 2023. Основное отличие от конкурсной версии — гораздо более совершенные вокальные модели, которые здесь используются. Для вокала мы используем следующие 4 модели: UVR-MDX-NET-Voc_FT, Demucs4 Vocals 2023, лучшую модель MDX23C и VitLarge23. Для дорожек «bass», «drums» и «other» мы используем следующие 4 модели: demucsht_ft, demucs_ht, demucs_6s и demucs_mmi. Первоначальная конкурсная модель доступна здесь: https://github.com/ZFTurbo/MVSEP-MDX23-music-separation-model

\r\n

Таблицы качества

\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n
Multisong datasetSynth dataset
SDR BassSDR DrumsSDR OtherSDR VocalsSDR InstrumentalSDR VocalsSDR Instrumental
12.5211.737.0110.3016.6012.6712.38
\r\n

 

\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n
MDX23 Leaderboard
SDR BassSDR DrumsSDR OtherSDR Vocals
9.9379.5597.28011.093
", "lang": "ru", "algorithm_id": 22, "created_at": "2023-08-09 18:12:12", "updated_at": "2023-10-04 03:07:31" } ] }, { "id": 32, "name": "Ensemble All-In (vocals, bass, drums, piano, guitar, lead/back vocals, other)", "algorithm_group_id": 3, "orientation": 2, "render_id": 30, "order_id": 3, "created_at": "2023-10-03 17:14:37", "updated_at": "2023-10-03 17:14:37", "additional_fields": null, "price_coefficient": 6, "is_active": 1, "algorithm_fields": [ { "id": 20, "name": "add_opt1", "text": "Output files", "algorithm_id": 32, "options": "{\n \"0\": \"Standard set\",\n \"1\": \"Include intermediate results and max_fft, min_fft\"\n}", "created_at": "2024-01-03 23:26:56", "updated_at": "2024-01-03 23:26:56", "default_key": "0", "server_key": "detailed" } ], "algorithm_descriptions": [] }, { "id": 21, "name": "Demucs4 HT (vocals, drums, bass, other)", "algorithm_group_id": 2, "orientation": 1, "render_id": 20, "order_id": 1, "created_at": "2022-11-28 09:37:27", "updated_at": "2023-03-13 02:52:03", "additional_fields": null, "price_coefficient": 1, "is_active": 1, "algorithm_fields": [ { "id": 7, "name": "add_opt1", "text": "Model type", "algorithm_id": 21, "options": "{\n \"0\": \"htdemucs_ft (High Quality, Slow)\",\n \"1\": \"htdemucs (Good Quality, Fast)\",\n \"2\": \"htdemucs_6s (6 stems, additional piano and guitar)\"\n}", "created_at": "2023-02-28 17:13:55", "updated_at": "2023-06-12 12:25:27", "default_key": "1", "server_key": "model_type" } ], "algorithm_descriptions": [ { "id": 3, "short_description": "Algorithm Demucs4 HT. Best for bass/drums/other.", "long_description": "

Algorithm Demucs4 HT splits track into 4 stems (bass, drums, vocals, other). It's now best for bass/drums/other separation. It was released in year 2022 and has 3 versions: 

\r\n\r\n

Link: https://github.com/facebookresearch/demucs/tree/ht/demucs

\r\n

Quality table

\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n
Algorithm nameMultisong datasetSynth datasetMDX23 Leaderboard
SDR BassSDR DrumsSDR OtherSDR VocalsSDR InstrumentalSDR VocalsSDR InstrumentalSDR Vocals
htdemucs_ft12.0511.245.748.3314.6310.239.949.08
htdemucs11.7410.905.578.1814.49------ 
htdemucs_6d11.4210.592.638.1714.48------ 
", "lang": "en", "algorithm_id": 21, "created_at": "2023-04-08 14:44:31", "updated_at": "2023-08-01 15:02:20" }, { "id": 4, "short_description": "Алгоритм Demucs4 HT. Лучший для разделения на bass/drums/other.", "long_description": "

Алгоритм Demucs4 HT. Лучший для разделения на bass/drums/other. Он был представлен в 2022 году командой из Facebook research. У алгоритма 3 версии: 

\r\n\r\n

Ссылка: https://github.com/facebookresearch/demucs/tree/ht/demucs

\r\n

Метрики качества разделения

\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n
Название алгоритмаMultisong datasetSynth datasetMDX23 Leaderboard
SDR BassSDR DrumsSDR OtherSDR VocalsSDR InstrumentalSDR VocalsSDR InstrumentalSDR Vocals
htdemucs_ft12.0511.245.748.3314.6310.239.949.08
htdemucs11.7410.905.578.1814.49------ 
htdemucs_6d11.4210.592.638.1714.48------ 
", "lang": "ru", "algorithm_id": 21, "created_at": "2023-04-08 15:16:15", "updated_at": "2023-08-01 15:04:04" } ] }, { "id": 26, "name": "MDX23C (vocals, instrumental)", "algorithm_group_id": 2, "orientation": 0, "render_id": 25, "order_id": 2, "created_at": "2023-07-05 15:32:54", "updated_at": "2023-07-05 15:33:54", "additional_fields": null, "price_coefficient": 1, "is_active": 1, "algorithm_fields": [ { "id": 10, "name": "add_opt1", "text": "Vocal model type", "algorithm_id": 26, "options": "{\n \"3\": \"12K FFT, Large Conv, Hop 1024 (SDR vocals: 9.95, SDR instrum: 16.26)\",\n \"2\": \"12K FFT, Large Conv (SDR vocals: 9.71, SDR instrum: 16.02)\",\n \"0\": \"12K FFT (SDR vocals: 9.68, SDR instrum: 15.99)\",\n \"1\": \"12K FFT, 6 Poolings (SDR vocals: 9.49, SDR instrum: 15.79)\",\n \"4\": \"8K FFT (SDR vocals: 10.17, SDR instrum: 16.48)\",\n \"7\": \"8K FFT (SDR vocals: 10.36, SDR instrum: 16.66)\"\n}", "created_at": "2023-07-05 15:39:05", "updated_at": "2023-11-10 03:10:51", "default_key": "7", "server_key": "model_type" } ], "algorithm_descriptions": [ { "id": 7, "short_description": "MDX23C models. Best for vocals/instrumental separation.", "long_description": "

New set of models MDX23C is based on code released by kuielab for Sound Demixing Challenge 2023. All models are full band, e.g. they don't cut high frequences.

\r\n

Quality table

\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n
Algorithm nameMultisong datasetSynth datasetMDX23 Leaderboard
SDR VocalsSDR InstrumentalSDR VocalsSDR InstrumentalSDR Vocals
12K FFT, Full Band, Large Conv, Hop 10249.9516.2611.7411.4410.78
12K FFT, Full Band, Large Conv9.7116.02---------
12K FFT, Full Band9.6815.99---------
12K FFT, Full Band, 6 Poolings9.4915.79---------
8K FFT, Full Band (v1)10.1716.4812.3512.0611.04
8K FFT, Full Band (v2)10.3616.6612.5212.2211.16
", "lang": "en", "algorithm_id": 26, "created_at": "2023-07-05 22:52:12", "updated_at": "2023-11-20 16:56:09" }, { "id": 8, "short_description": "Модели MDX23C. Лучший выбор для разделения на вокальную и инструментальную дорожки.", "long_description": "

Новый набор моделей MDX23C, основанный на коде опубликованном kuielab в рамках Sound Demixing Challenge 2023. Результаты полученных моделей содержат весь частотный спектр и имеют максимальные метрики качества для вокала и музыки на MultiSong Dataset и Synth Dataset.

\r\n

Таблица качества

\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n
Algorithm nameMultisong datasetSynth datasetMDX23 Leaderboard
SDR VocalsSDR InstrumentalSDR VocalsSDR InstrumentalSDR Vocals
12K FFT, Full Band, Large Conv, Hop 10249.9516.2611.7411.4410.78
12K FFT, Full Band, Large Conv9.7116.02---------
12K FFT, Full Band9.6815.99---------
12K FFT, Full Band, 6 Poolings9.4915.79---------
8K FFT, Full Band10.1716.4812.3512.0611.04
8K FFT, Full Band (v2)10.3616.6612.5212.2211.16
", "lang": "ru", "algorithm_id": 26, "created_at": "2023-07-05 23:05:50", "updated_at": "2023-11-20 16:56:28" } ] }, { "id": 24, "name": "MDX B (vocals, instrumental)", "algorithm_group_id": 2, "orientation": 0, "render_id": 23, "order_id": 3, "created_at": "2023-05-21 13:17:51", "updated_at": "2023-07-05 15:34:06", "additional_fields": null, "price_coefficient": 1, "is_active": 1, "algorithm_fields": [ { "id": 8, "name": "add_opt1", "text": "Vocal model type", "algorithm_id": 24, "options": "{\n \"7\": \"MDX Kimberley Jensen v2 2023.05.21 (SDR: 9.60)\",\n \"0\": \"MDX UVR 2022.01.01 (SDR: 8.83)\",\n \"1\": \"MDX UVR 2022.07.25(SDR: 8.67)\",\n \"2\": \"MDX Kimberley Jensen v1 2023.02.12 (SDR: 9.48)\",\n \"4\": \"UVR-MDX-NET-Inst_HQ_2 (SDR: 9.12)\",\n \"5\": \"UVR_MDXNET_Main (SDR: 8.79)\",\n \"6\": \"MDX Kimberley Jensen Inst (SDR: 9.28)\",\n \"8\": \"UVR-MDX-NET-Inst_HQ_3 (SDR: 9.38)\",\n \"9\": \"UVR-MDX-NET-Voc_FT (SDR: 9.64)\"\n}", "created_at": "2023-05-21 13:23:46", "updated_at": "2023-07-31 01:32:52", "default_key": "9", "server_key": "model_type" } ], "algorithm_descriptions": [ { "id": 12, "short_description": "MDX B models are based on kuielab code from Music Demixing Challenge 2021. Models were retrained by UVR team on big dataset. For long time models were best for vocals/instrumental separation.", "long_description": "

MDX B models are based on kuielab code from Music Demixing Challenge 2021. Models were retrained by UVR team on big dataset. For long time models were best for vocals/instrumental separation.

\r\n

Quality table

\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n
Algorithm nameMultisong datasetSynth datasetMDX23 Leaderboard
SDR VocalsSDR InstrumentalSDR VocalsSDR InstrumentalSDR Vocals
UVR-MDX-NET-Voc_FT9.6415.9511.4011.1010.505
MDX Kimberley Jensen v29.6015.91------10.494
MDX Kimberley Jensen v19.4815.79---------
UVR-MDX-NET-Inst_HQ_39.3815.6811.3211.0310.254
MDX Kimberley Jensen Inst9.2815.59---------
UVR-MDX-NET-Inst_HQ_29.1215.42---------
MDX UVR 2022.01.018.8315.14---------
UVR_MDXNET_Main8.7915.10---------
MDX UVR 2022.07.258.6714.97---------
", "lang": "en", "algorithm_id": 24, "created_at": "2023-08-03 04:00:50", "updated_at": "2023-08-03 14:41:43" } ] }, { "id": 3, "name": "Ultimate Vocal Remover HQ (vocals, music)", "algorithm_group_id": 2, "orientation": 1, "render_id": 9, "order_id": 4, "created_at": null, "updated_at": "2023-07-05 15:33:38", "additional_fields": null, "price_coefficient": 1, "is_active": 1, "algorithm_fields": [ { "id": 2, "name": "add_opt1", "text": "Model Type", "algorithm_id": 3, "options": "{\n \"0\": \"HP2-4BAND-3090_4band_arch-500m_1\",\n \"1\": \"HP2-4BAND-3090_4band_2\",\n \"2\": \"HP2-4BAND-3090_4band_1\",\n \"3\": \"HP_4BAND_3090\",\n \"4\": \"Vocal_HP_4BAND_3090\",\n \"5\": \"Vocal_HP_4BAND_3090_AGG\",\n \"6\": \"HP2-MAIN-MSB2-3BAND-3090\",\n \"7\": \"HP-4BAND-V2\",\n \"8\": \"HP-KAROKEE-MSB2-3BAND-3090 (Karaokee model)\",\n \"9\": \"WIP-Piano-4band-129605kb (Piano model)\",\n \"10\": \"drums-4BAND-3090_4band (Drums model)\",\n \"11\": \"bass-4BAND-3090_4band (Bass model)\",\n \"12\": \"karokee_4band_v2_sn (Karaokee model v2)\",\n \"13\": \"UVR-De-Echo-Aggressive\",\n \"14\": \"UVR-De-Echo-Normal\",\n \"15\": \"UVR-DeNoise\",\n \"16\": \"UVR-DeEcho-DeReverb\",\n \"17\": \"UVR-BVE-4B_SN-44100-1 (Back vocals model)\"\n}", "created_at": "2022-12-07 09:33:55", "updated_at": "2023-09-23 16:40:15", "default_key": "0", "server_key": "model_type" }, { "id": 3, "name": "add_opt2", "text": "Agressiveness", "algorithm_id": 3, "options": "{\n \"0.3\": \"0.3\",\n \"0.1\": \"0.1\",\n \"0.2\": \"0.2\",\n \"0.4\": \"0.4\",\n \"0.5\": \"0.5\",\n \"0.6\": \"0.6\",\n \"0.7\": \"0.7\",\n \"0.8\": \"0.8\",\n \"0.9\": \"0.9\",\n \"1.0\": \"1.0\"\n}", "created_at": "2022-12-07 09:43:18", "updated_at": "2023-03-12 02:08:07", "default_key": null, "server_key": "agg" } ], "algorithm_descriptions": [] }, { "id": 28, "name": "Demucs4 Vocals 2023 (vocals, instrum)", "algorithm_group_id": 2, "orientation": 0, "render_id": 27, "order_id": 5, "created_at": "2023-08-01 14:35:45", "updated_at": "2023-08-01 14:35:45", "additional_fields": null, "price_coefficient": 1, "is_active": 1, "algorithm_fields": [], "algorithm_descriptions": [ { "id": 11, "short_description": "Demucs4 Vocals 2023 model - it's Demucs4 HT model fine-tuned on big vocals dataset.", "long_description": "

Demucs4 Vocals 2023 model - it's Demucs4 HT model fine-tuned on big vocal/instrumental dataset. It has better metrics for vocals separation compared to Demucs4 HT (_ft version). It usually gives worse metrics than MDX23C models, but can be useful for ensembles, since the model is very different from MDX23C. 

\r\n

Quality table

\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n
Algorithm nameMultisong datasetSynth datasetMDX23 Leaderboard
SDR VocalsSDR InstrumentalSDR VocalsSDR InstrumentalSDR Vocals
Demucs4 Vocals 20239.0415.3511.5911.299.61
", "lang": "en", "algorithm_id": 28, "created_at": "2023-08-01 14:51:16", "updated_at": "2023-08-01 15:30:42" } ] }, { "id": 7, "name": "MDX-B Karaoke (lead/back vocals)", "algorithm_group_id": 2, "orientation": 1, "render_id": 12, "order_id": 7, "created_at": null, "updated_at": "2023-09-18 18:05:16", "additional_fields": null, "price_coefficient": 1, "is_active": 1, "algorithm_fields": [ { "id": 13, "name": "add_opt1", "text": "Karaoke model type", "algorithm_id": 7, "options": "{\n \"0\": \"Extract directly from mixture (SDR lead vocals: 6.81)\",\n \"1\": \"Extract from vocals part (SDR lead vocals: 7.94)\"\n}", "created_at": "2023-09-18 15:48:30", "updated_at": "2023-09-18 15:48:30", "default_key": "0", "server_key": "model_type" } ], "algorithm_descriptions": [ { "id": 18, "short_description": "Модель MDX-B Karaoke подготовлена в рамках проекта Ultimate Vocal Remover. Модель производит высококачественное извлечение ведущей вокальной партии из музыкального трека.", "long_description": "

Модель MDX-B Karaoke подготовлена в рамках проекта Ultimate Vocal Remover. Модель производит высококачественное извлечение ведущей вокальной партии из музыкального трека. Модель доступна в двух вариантах. В первом варианте, нейросетевая модель используется напрямую на всем треке. Во втором случае, трек сначала разделяется на две части вокальную и инструментальную и затем нейросетевая модель применяется уже только к вокальной части. Во втором случае качество разделения обычно выше и при этом появляется возможность дополнительно отделить бэк-вокал в отдельную дорожку. Модель сравнивалась с двумя другими моделями из UVR (они тоже доступны на сайте) на большом валидационном наборе. Используемая метрика — SDR: чем больше, тем лучше.

\r\n

См. результаты в таблице ниже.

\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n
Тип валидацииНазвание алгоритма
UVR (HP-KAROKEE-MSB2-3BAND-3090)UVR (karokee_4band_v2_sn)MDX-B Karaoke (Type 0)MDX-B Karaoke (Type 1)
Validation lead vocals6.466.346.817.94
Validation other13.1713.0213.5314.66
Validation back vocals---------1.88
", "lang": "ru", "algorithm_id": 7, "created_at": "2023-09-18 17:06:51", "updated_at": "2023-09-18 22:38:54" }, { "id": 19, "short_description": "The MDX-B Karaoke model was prepared as part of the Ultimate Vocal Remover project. The model produces high-quality lead vocal extraction from a music track.", "long_description": "

The MDX-B Karaoke model was prepared as part of the Ultimate Vocal Remover project. The model produces high-quality lead vocal extraction from a music track. The model is available in two versions. In the first version, the neural network model is used directly on the entire track. In the second version, the track is first divided into two parts, vocal and instrumental, and then the neural network model is applied only to the vocal part. In the second version, the quality of separation is usually higher and it becomes possible to additionally separate the backing vocals into a separate track. The model was compared with two other models from UVR (they are also available on the website) on a large validation set. The metric used is SDR: the higher the better.

\r\n

See the results in the table below.

\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n
Validation typeAlgorithm name
UVR (HP-KAROKEE-MSB2-3BAND-3090)UVR (karokee_4band_v2_sn)MDX-B Karaoke (Type 0)MDX-B Karaoke (Type 1)
Validation lead vocals6.466.346.817.94
Validation other13.1713.0213.5314.66
Validation back vocals---------1.88
\r\n

 

", "lang": "en", "algorithm_id": 7, "created_at": "2023-09-18 17:19:32", "updated_at": "2023-09-18 22:39:02" } ] }, { "id": 23, "name": "FoxJoy Reverb Removal (other)", "algorithm_group_id": 2, "orientation": 0, "render_id": 22, "order_id": 13, "created_at": "2023-04-25 04:12:28", "updated_at": "2023-04-25 04:34:50", "additional_fields": null, "price_coefficient": 1, "is_active": 1, "algorithm_fields": [], "algorithm_descriptions": [] }, { "id": 25, "name": "MVSep Demucs4HT DNR (dialog, sfx, music)", "algorithm_group_id": 2, "orientation": 0, "render_id": 24, "order_id": 14, "created_at": "2023-06-09 02:55:39", "updated_at": "2023-06-09 02:55:39", "additional_fields": null, "price_coefficient": 1, "is_active": 1, "algorithm_fields": [ { "id": 9, "name": "add_opt1", "text": "Model type", "algorithm_id": 25, "options": "{\n \"0\": \"Single (SDR: 9.62)\",\n \"1\": \"Ensemble (SDR: 10.16)\"\n}", "created_at": "2023-06-09 02:58:00", "updated_at": "2023-06-09 02:58:00", "default_key": "0", "server_key": "model_type" } ], "algorithm_descriptions": [] }, { "id": 35, "name": "BandIt Plus (speech, music, effects)", "algorithm_group_id": 2, "orientation": 0, "render_id": 36, "order_id": 36, "created_at": "2023-12-21 03:34:08", "updated_at": "2023-12-21 03:34:08", "additional_fields": null, "price_coefficient": 1, "is_active": 1, "algorithm_fields": [], "algorithm_descriptions": [ { "id": 26, "short_description": "Модель BandIt Plus для разделения треков на голос, музыку и эффекты.", "long_description": "

Модель BandIt Plus для разделения треков на голос, музыку и эффекты. Она может быть полезна для телевизионных или кинематографических роликов. Модель была выложена авторами статьи \"A Generalized Bandsplit Neural Network for Cinematic Audio Source Separation\" в репозитории на GitHub. Модель была натренирована на наборе данных Divide and Remaster (DnR). И на данный момент имеет лучшие метрики качества среди подобных моделей.

\r\n

Таблица качества

\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n
Название алгоритмаDnR dataset
SDR SpeechSDR MusicSDR Effects
BandIt Plus15.629.219.69
", "lang": "ru", "algorithm_id": 35, "created_at": "2023-12-21 13:49:03", "updated_at": "2023-12-21 13:51:22" }, { "id": 27, "short_description": "BandIt Plus model for separating tracks into speech, music and effects.", "long_description": "

BandIt Plus model for separating tracks into speech, music and effects. The model can be useful for television or film clips. The model was prepared by the authors of the article \"A Generalized Bandsplit Neural Network for Cinematic Audio Source Separation\" in the repository on GitHub. The model was trained on the Divide and Remaster (DnR) dataset. And at the moment it has the best quality metrics among similar models.

\r\n

Quality table

\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n
Algorithm nameDnR dataset (test)
SDR SpeechSDR MusicSDR Effects
BandIt Plus15.629.219.69
", "lang": "en", "algorithm_id": 35, "created_at": "2023-12-21 13:51:04", "updated_at": "2023-12-21 13:51:04" } ] }, { "id": 29, "name": "MVSep Piano (piano, other)", "algorithm_group_id": 2, "orientation": 0, "render_id": 29, "order_id": 100, "created_at": "2023-08-16 17:50:33", "updated_at": "2023-08-16 17:50:33", "additional_fields": null, "price_coefficient": 1, "is_active": 1, "algorithm_fields": [ { "id": 11, "name": "add_opt1", "text": "Piano model type", "algorithm_id": 29, "options": "{\n \"0\": \"Extract directly from mixture (SDR piano: 7.26)\",\n \"1\": \"Extract from instrumental part (SDR piano: 7.31)\"\n}", "created_at": "2023-09-18 02:08:43", "updated_at": "2023-09-18 02:08:43", "default_key": "0", "server_key": "model_type" } ], "algorithm_descriptions": [ { "id": 14, "short_description": "MVSep Piano model is based on MDX23C architecture. It produces high quality separation for piano and other stems.", "long_description": "

MVSep Piano model is based on MDX23C architecture. It produces high quality separation. Model was compared with other two models (Demucs4HT (6 stems) and GSEP) on two validation sets. First validation includes electric piano as part of piano, while 2nd only contains acoustic piano (grand piano). Used metrics is SDR: the more the better.

\r\n

See the results in table below. 

\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n
Validation typeAlgorithm name
Demucs4HT (6 stems)GSEPMVSep Piano 2023 (Type 0)MVSep Piano 2023 (Type 1)
Validation full2.44323.55894.91874.9772
Validation (only grand piano)4.55915.71807.26517.2948
\r\n

The model is available in two variants. In the first variant, the Piano model is used directly on the entire track. In the second variant, the track is first divided into two parts, vocal and instrument, and then the Piano model is applied to the instrument part only. In the second case, the separation quality is usually a bit better.

", "lang": "en", "algorithm_id": 29, "created_at": "2023-08-16 17:58:19", "updated_at": "2023-09-18 02:41:03" }, { "id": 15, "short_description": "Модель MVSep Piano основана на архитектуре MDX23C. Модель производит высококачественное разделение музыки на партию фортепиано и всего остального.", "long_description": "

Модель MVSep Piano основана на архитектуре MDX23C. Она обеспечивает высокое качество разделения. Модель сравнивалась с двумя другими моделями (Demucs4HT (6 партий) и GSEP) на двух валидационных наборах. Первый набор включает электрическое пианино как часть партии фортепиано, а вторая включает только акустическое пианино (рояль). Используемая метрика — SDR: чем больше, тем лучше.

\r\n

См. результаты в таблице ниже.

\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n
Тип валидацииНазвание алгоритма
Demucs4HT (6 stems)GSEPMVSep Piano 2023 (Type 0)MVSep Piano 2023 (Type 1)
Validation full2.44323.55894.91874.9772
Validation (only grand piano)4.55915.71807.26517.2948
\r\n

Модель доступна в двух вариантах. В первом варианте, модель пианино используется напрямую на всем треке. Во втором случае, трек сначала разделяется на две части вокальную и инструментальную и затем модель пианино применяется уже только к инструментальной части. Во втором случае качество разделения обычно немного выше.

", "lang": "ru", "algorithm_id": 29, "created_at": "2023-08-16 17:58:43", "updated_at": "2023-09-18 13:36:27" } ] }, { "id": 30, "name": "MVSep Guitar (guitar, other)", "algorithm_group_id": 2, "orientation": 0, "render_id": 31, "order_id": 101, "created_at": "2023-09-18 13:10:19", "updated_at": "2023-09-18 13:10:19", "additional_fields": null, "price_coefficient": 1, "is_active": 1, "algorithm_fields": [ { "id": 12, "name": "add_opt1", "text": "Guitar model type", "algorithm_id": 30, "options": "{\n \"0\": \"Extract directly from mixture (SDR guitar: 7.77)\",\n \"1\": \"Extract from instrumental part (SDR guitar: 7.92)\"\n}", "created_at": "2023-09-18 13:11:42", "updated_at": "2023-09-18 13:11:42", "default_key": "0", "server_key": "model_type" } ], "algorithm_descriptions": [ { "id": 16, "short_description": "Модель MVSep Guitar основана на архитектуре MDX23C. Модель производит высококачественное разделение музыки на партию гитары (включает акустическую и электронную) и всего остального.", "long_description": "

Модель MVSep Guitar основана на архитектуре MDX23C. Модель производит высококачественное разделение музыки на партию гитары (включает акустическую и электронную) и всего остального. Модель сравнивалась с моделью Demucs4HT (6 партий) на валидационном наборе для гитары. Используемая метрика — SDR: чем больше, тем лучше.

\r\n

См. результаты в таблице ниже.

\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n
Тип валидацииНазвание алгоритма
Demucs4HT (6 stems)MVSep Guitar 2023 (Type 0)MVSep Guitar 2023 (Type 1)
Validation guitar7.22457.77167.9251
Validation other13.175613.722713.8762
\r\n

Модель доступна в двух вариантах. В первом варианте, нейросетевая модель для гитары используется напрямую на всем треке. Во втором случае, трек сначала разделяется на две части вокальную и инструментальную и затем нейросетевая модель для гитары применяется уже только к инструментальной части. Во втором случае качество разделения обычно немного выше.

", "lang": "ru", "algorithm_id": 30, "created_at": "2023-09-18 13:34:46", "updated_at": "2023-09-18 13:34:52" }, { "id": 17, "short_description": "The MVSep Guitar model is based on the MDX23C architecture. The model produces high-quality separation of music into a guitar part (including acoustic and electronic) and everything else.", "long_description": "

The MVSep Guitar model is based on the MDX23C architecture. The model produces high-quality separation of music into a guitar part (including acoustic and electronic) and everything else. The model was compared with the Demucs4HT model (6 stems) on a guitar validation set. The metric used is SDR: the higher the better.

\r\n

See the results in the table below.

\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n
Validation typeAlgorithm name
Demucs4HT (6 stems)MVSep Guitar 2023 (Type 0)MVSep Guitar 2023 (Type 1)
Validation guitar7.22457.77167.9251
Validation other13.175613.722713.8762
\r\n

The model is available in two versions. In the first version, the neural network model for the guitar is used directly on the entire track. In the second case, the track is first divided into two parts, vocal and instrumental, and then the neural network model for the guitar is applied only to the instrumental part. In the second case, the separation quality is usually slightly higher.

", "lang": "en", "algorithm_id": 30, "created_at": "2023-09-18 13:35:11", "updated_at": "2023-09-18 13:40:06" } ] }, { "id": 31, "name": "Vit Large 23 (vocals, instrum)", "algorithm_group_id": 2, "orientation": 0, "render_id": 33, "order_id": 102, "created_at": "2023-09-29 14:16:07", "updated_at": "2023-09-29 14:16:18", "additional_fields": null, "price_coefficient": 1, "is_active": 1, "algorithm_fields": [ { "id": 14, "name": "add_opt1", "text": "Model type", "algorithm_id": 31, "options": "{\n \"0\": \"v1 (SDR vocals: 9.78)\",\n \"1\": \"v2 (SDR vocals: 9.90)\"\n}", "created_at": "2023-12-01 13:57:03", "updated_at": "2023-12-01 13:57:03", "default_key": "1", "server_key": "model_type" } ], "algorithm_descriptions": [ { "id": 20, "short_description": "Экспериментальная модель VitLarge23 на базе Vision Transformers. По метрикам немного уступает MDX23C, но может сработать лучше в некоторых случаях.", "long_description": "

Экспериментальная модель VitLarge23 на базе Vision Transformers. По метрикам немного уступает MDX23C, но может сработать лучше в некоторых случаях.

\r\n

Таблица качества

\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n
Algorithm nameMultisong datasetSynth datasetMDX23 Leaderboard
SDR VocalsSDR InstrumentalSDR VocalsSDR InstrumentalSDR Vocals
Vit Large 23 (512px) v19.7816.0912.3312.0310.47 
Vit Large 23 (512px) v29.9016.2012.3812.08---
", "lang": "ru", "algorithm_id": 31, "created_at": "2023-09-30 19:53:23", "updated_at": "2023-12-01 14:03:12" }, { "id": 21, "short_description": "Experimental model VitLarge23 based on Vision Transformers. In terms of metrics, it is slightly inferior to the MDX23C, but may work better in some cases.", "long_description": "

Experimental model VitLarge23 based on Vision Transformers. In terms of metrics, it is slightly inferior to the MDX23C, but may work better in some cases.

\r\n

Quality table

\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n
Algorithm nameMultisong datasetSynth datasetMDX23 Leaderboard
SDR VocalsSDR InstrumentalSDR VocalsSDR InstrumentalSDR Vocals
Vit Large 23 (512px) v19.7816.0912.3312.0310.47 
Vit Large 23 (512px) v29.9016.2012.3812.08---
", "lang": "en", "algorithm_id": 31, "created_at": "2023-09-30 19:53:58", "updated_at": "2023-12-01 14:02:47" } ] }, { "id": 33, "name": "MVSep Crowd removal (crowd, other)", "algorithm_group_id": 2, "orientation": 0, "render_id": 34, "order_id": 200, "created_at": "2023-11-20 00:20:35", "updated_at": "2023-11-20 00:21:40", "additional_fields": null, "price_coefficient": 1, "is_active": 1, "algorithm_fields": [ { "id": 15, "name": "add_opt1", "text": "Model type", "algorithm_id": 33, "options": "{\n \"8\": \"v1 (SDR crowd: 5.57)\",\n \"9\": \"v2 (SDR crowd: 6.06)\"\n}", "created_at": "2023-12-09 02:44:02", "updated_at": "2023-12-09 14:55:58", "default_key": "9", "server_key": "model_type" } ], "algorithm_descriptions": [ { "id": 24, "short_description": "An unique model for removing crowd sounds from music recordings (applause, clapping, whistling, noise, laugh etc.).", "long_description": "

An unique model for removing crowd sounds from music recordings (applause, clapping, whistling, noise, laugh etc.). Current metrics on our internal dataset for quality control:

\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n
Algorithm nameCrowd dataset
SDR crowdSDR other
Crowd model MDX23C (v1)5.5718.79
Crowd model MDX23C (v2)6.0619.28
\r\n

Examples of how the model works can be found: here и here.

", "lang": "en", "algorithm_id": 33, "created_at": "2023-12-09 14:59:06", "updated_at": "2023-12-09 15:15:02" }, { "id": 25, "short_description": "Уникальная модель для удаления звуков толпы из музыкальных записей (аплодисменты, хлопки, свист, шум, смех и т.д.).", "long_description": "

Уникальная модель для удаления звуков толпы из музыкальных записей (аплодисменты, хлопки, свист, шум, смех и т.д.). Текущие показатели нашего внутреннего набора данных для контроля качества:

\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n
Algorithm nameCrowd dataset
SDR crowdSDR other
Crowd model MDX23C (v1)5.5718.79
Crowd model MDX23C (v2)6.0619.28
\r\n

Примеры работы модели можно найти: здесь и здесь.

", "lang": "ru", "algorithm_id": 33, "created_at": "2023-12-09 15:00:21", "updated_at": "2023-12-09 15:14:45" } ] }, { "id": 34, "name": "MVSep MelBand Roformer (vocals, instrum)", "algorithm_group_id": 2, "orientation": 0, "render_id": 35, "order_id": 300, "created_at": "2023-12-02 16:33:45", "updated_at": "2023-12-02 16:34:18", "additional_fields": null, "price_coefficient": 1, "is_active": 1, "algorithm_fields": [], "algorithm_descriptions": [ { "id": 22, "short_description": "Mel Band Roformer - модель предложенная победителями конкурса Sound Demixing Challenge 2023", "long_description": "

Mel Band Roformer - модель предложенная сотрудниками компании ByteDance для конкурса Sound Demixing Challenge 2023, где они заняли первое место на LeaderBoard C. К сожалению, модель не была выложена в открытый доступ и была воспроизведена по научной статье разработчиком @lucidrains на площадке github. Вокальная модель была обучена с нуля на нашем внутреннем датасете. К сожалению, нам пока не удалось достичь аналогичных метрик как у авторов.

\r\n

Таблица качества

\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n
Algorithm nameMultisong datasetSynth datasetMDX23 Leaderboard
SDR VocalsSDR InstrumentalSDR VocalsSDR InstrumentalSDR Vocals
Mel Band Roformer v1 (vocals)9.07---11.76------
", "lang": "ru", "algorithm_id": 34, "created_at": "2023-12-02 17:04:22", "updated_at": "2023-12-02 17:09:38" }, { "id": 23, "short_description": "Mel Band Roformer - model proposed by the winners of the Sound Demixing Challenge 2023", "long_description": "

Mel Band Roformer - a model proposed by employees of the company ByteDance for the competition Sound Demixing Challenge 2023, where they took first place on LeaderBoard C. Unfortunately, the model was not made publicly available and was reproduced according to a scientific article by the developer @lucidrains on the github. The vocal model was trained from scratch on our internal dataset. Unfortunately, we have not yet been able to achieve similar metrics as the authors.

\r\n

Quality table

\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n
Algorithm nameMultisong datasetSynth datasetMDX23 Leaderboard
SDR VocalsSDR InstrumentalSDR VocalsSDR InstrumentalSDR Vocals
Mel Band Roformer v1 (vocals)9.07---11.76------
", "lang": "en", "algorithm_id": 34, "created_at": "2023-12-02 17:10:56", "updated_at": "2023-12-02 17:10:56" } ] }, { "id": 36, "name": "DrumSep (kick, snare, cymbals, toms)", "algorithm_group_id": 2, "orientation": 0, "render_id": 37, "order_id": 400, "created_at": "2023-12-24 16:43:42", "updated_at": "2023-12-24 16:45:03", "additional_fields": null, "price_coefficient": 1, "is_active": 1, "algorithm_fields": [ { "id": 16, "name": "add_opt1", "text": "Model type", "algorithm_id": 36, "options": "{\n \"0\": \"Apply Demucs4HT first to get drums\",\n \"1\": \"Use as is (audio must contain drums only)\"\n}", "created_at": "2023-12-24 18:20:28", "updated_at": "2023-12-24 18:20:28", "default_key": "0", "server_key": "model_type" } ], "algorithm_descriptions": [ { "id": 28, "short_description": "Модель производит разделение дорожки барабанов на 4 типа: 'kick', 'snare', 'cymbals', 'toms'.", "long_description": "

Модель производит разделение дорожки барабанов на 4 типа: 'kick', 'snare', 'cymbals', 'toms'. Используется модель DrumSep из репозитория на github. Модель имеет два режима работы. В первом (по умолчанию) сначала к треку применяется модель Demucs4 HT, которая извлекает из трека только барабанную часть. Далее уже применяется модель DrumSep. Если ваш трек состоит только из барабанов, то имеет смысл использовать второй режим, где модель DrumSep применяется непосредственно к загруженному аудио.

", "lang": "ru", "algorithm_id": 36, "created_at": "2023-12-24 18:25:23", "updated_at": "2023-12-24 18:25:23" }, { "id": 29, "short_description": "The DrumSep model divides the drums stem into 4 types: 'kick', 'snare', 'cymbals', 'toms'.", "long_description": "

The DrumSep model divides the drums stem into 4 types: 'kick', 'snare', 'cymbals', 'toms'. The model from this github repository is used. The model has two operating modes. The first (default) first applies the Demucs4 HT model to the track, which extracts only the drums part from the track. Next, the DrumSep model is applied. If your track consists only of drums, then it makes sense to use the second mode, where the DrumSep model is applied directly to the uploaded audio.

", "lang": "en", "algorithm_id": 36, "created_at": "2023-12-24 18:28:13", "updated_at": "2023-12-24 18:28:13" } ] }, { "id": 37, "name": "LarsNet (kick, snare, cymbals, toms, hihat)", "algorithm_group_id": 2, "orientation": 0, "render_id": 38, "order_id": 500, "created_at": "2023-12-25 16:10:53", "updated_at": "2023-12-25 16:10:53", "additional_fields": null, "price_coefficient": 1, "is_active": 1, "algorithm_fields": [ { "id": 17, "name": "add_opt1", "text": "Model type", "algorithm_id": 37, "options": "{\n \"0\": \"Apply Demucs4HT first to get drums\",\n \"1\": \"Use as is (audio must contain drums only)\"\n}", "created_at": "2023-12-25 16:11:39", "updated_at": "2023-12-25 16:11:39", "default_key": "0", "server_key": "model_type" } ], "algorithm_descriptions": [ { "id": 30, "short_description": "Модель LarsNet производит разделение дорожки барабанов на 5 типов: 'kick', 'snare', 'cymbals', 'toms', 'hihat'.", "long_description": "

Модель LarsNet производит разделение дорожки барабанов на 5 типов: 'kick', 'snare', 'cymbals', 'toms', 'hihat'. Используется модель из репозитория на github и обученная на датасете StemGMD. Модель имеет два режима работы. В первом (по умолчанию) сначала к треку применяется модель Demucs4 HT, которая извлекает из трека только барабанную часть. Далее уже применяется модель LarsNet. Если ваш трек состоит только из барабанов, то имеет смысл использовать второй режим, где модель LarsNet применяется непосредственно к загруженному аудио. К сожалению, субъективно качество разделения уступает по качеству модели DrumSep.

", "lang": "ru", "algorithm_id": 37, "created_at": "2023-12-26 03:05:19", "updated_at": "2023-12-26 03:05:19" }, { "id": 31, "short_description": "The LarsNet model divides the drums stem into 5 types: 'kick', 'snare', 'cymbals', 'toms', 'hihat'.", "long_description": "

The LarsNet model divides the drums stem into 5 types: 'kick', 'snare', 'cymbals', 'toms', 'hihat'. The model is from this github repository and it was trained on the dataset StemGMD. The model has two operating modes. The first (default) applies the Demucs4 HT model to the track at stage one, which extracts only the drum part from the track. On the second stage, the LarsNet model is used. If your track consists only of drums, then it makes sense to use the second mode, where the LarsNet model is applied directly to the uploaded audio. Unfortunately, subjectively, the quality of separation is inferior in quality to the model DrumSep.

", "lang": "en", "algorithm_id": 37, "created_at": "2023-12-26 03:07:58", "updated_at": "2023-12-26 03:07:58" } ] }, { "id": 38, "name": "test", "algorithm_group_id": 1, "orientation": 1, "render_id": 1, "order_id": 1, "created_at": "2024-01-19 12:34:51", "updated_at": "2024-01-19 12:34:51", "additional_fields": null, "price_coefficient": 1, "is_active": 1, "algorithm_fields": [], "algorithm_descriptions": [] }, { "id": 2, "name": "MDX B (vocals, drums, bass, other)", "algorithm_group_id": 1, "orientation": 1, "render_id": 8, "order_id": 2, "created_at": null, "updated_at": "2023-05-21 13:18:17", "additional_fields": null, "price_coefficient": 1, "is_active": 1, "algorithm_fields": [ { "id": 5, "name": "add_opt1", "text": "Vocal model type", "algorithm_id": 2, "options": "{\n \"2\": \"MDX Kimberley Jensen 2023.02.12 SDR: 9.30 (New)\",\n \"0\": \"MDX UVR 2022.01.01 SDR 8.62\",\n \"1\": \"MDX UVR 2022.07.25 SDR 8.51\"\n}", "created_at": "2022-12-07 11:08:31", "updated_at": "2023-03-12 02:05:28", "default_key": "2", "server_key": "model_type" } ], "algorithm_descriptions": [] }, { "id": 1, "name": "Demucs3 Model B (vocals, drums, bass, other)", "algorithm_group_id": 1, "orientation": 1, "render_id": 11, "order_id": 2, "created_at": null, "updated_at": "2023-05-21 13:19:31", "additional_fields": null, "price_coefficient": 1, "is_active": 1, "algorithm_fields": [], "algorithm_descriptions": [ { "id": 2, "short_description": "Algorithm Demucs3", "long_description": "

Algorithm Demucs3 splits track into 4 stems (bass, drums, vocals, other). The winner of the Music Demuxing Challenge 2021. 

\r\n

Link: https://github.com/facebookresearch/demucs/tree/v3

\r\n

Quality table

\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n
Algorithm nameMultisong datasetSynth dataset
SDR BassSDR DrumsSDR OtherSDR VocalsSDR InstrumentalSDR VocalsSDR Instrumental
demucs310.6910.275.358.1314.449.789.48
", "lang": "en", "algorithm_id": 1, "created_at": "2023-03-18 23:07:40", "updated_at": "2023-04-08 15:28:49" } ] }, { "id": 4, "name": "UVRv5 Demucs (vocals, music)", "algorithm_group_id": 1, "orientation": 1, "render_id": 17, "order_id": 4, "created_at": null, "updated_at": "2023-08-01 14:34:59", "additional_fields": null, "price_coefficient": 1, "is_active": 1, "algorithm_fields": [ { "id": 4, "name": "add_opt1", "text": "Model Type", "algorithm_id": 4, "options": "{\n \"0\": \"UVR_Demucs_Model_1\",\n \"1\": \"UVR_Demucs_Model_2\",\n \"2\": \"UVR_Demucs_Model_Bag\"\n}", "created_at": "2022-12-07 11:06:02", "updated_at": "2023-03-12 02:06:34", "default_key": null, "server_key": "model_type" } ], "algorithm_descriptions": [] }, { "id": 5, "name": "MVSep DNR (music, sfx, speech)", "algorithm_group_id": 1, "orientation": 1, "render_id": 18, "order_id": 5, "created_at": null, "updated_at": "2023-12-21 03:34:46", "additional_fields": null, "price_coefficient": 1, "is_active": 1, "algorithm_fields": [], "algorithm_descriptions": [] }, { "id": 6, "name": "MVSep Vocal Model (vocals, music)", "algorithm_group_id": 1, "orientation": 1, "render_id": 19, "order_id": 6, "created_at": null, "updated_at": "2023-08-01 14:36:27", "additional_fields": null, "price_coefficient": 1, "is_active": 1, "algorithm_fields": [], "algorithm_descriptions": [] }, { "id": 8, "name": "Demucs2 (vocals, drums, bass, other)", "algorithm_group_id": 1, "orientation": 0, "render_id": 13, "order_id": 8, "created_at": null, "updated_at": "2022-11-28 06:54:25", "additional_fields": null, "price_coefficient": 1, "is_active": 1, "algorithm_fields": [], "algorithm_descriptions": [] }, { "id": 9, "name": "Danna Sep (vocals, drums, bass, other)", "algorithm_group_id": 1, "orientation": 0, "render_id": 15, "order_id": 9, "created_at": null, "updated_at": "2022-11-28 06:54:16", "additional_fields": null, "price_coefficient": 1, "is_active": 1, "algorithm_fields": [], "algorithm_descriptions": [] }, { "id": 10, "name": "Byte Dance (vocals, drums, bass, other)", "algorithm_group_id": 1, "orientation": 0, "render_id": 16, "order_id": 10, "created_at": null, "updated_at": "2022-11-28 06:54:07", "additional_fields": null, "price_coefficient": 1, "is_active": 1, "algorithm_fields": [], "algorithm_descriptions": [] }, { "id": 11, "name": "spleeter (vocals, music)", "algorithm_group_id": 1, "orientation": 0, "render_id": 0, "order_id": 11, "created_at": null, "updated_at": null, "additional_fields": null, "price_coefficient": 1, "is_active": 1, "algorithm_fields": [], "algorithm_descriptions": [] }, { "id": 12, "name": "spleeter (vocals, drums, bass, other)", "algorithm_group_id": 1, "orientation": 0, "render_id": 1, "order_id": 12, "created_at": null, "updated_at": null, "additional_fields": null, "price_coefficient": 1, "is_active": 1, "algorithm_fields": [], "algorithm_descriptions": [] }, { "id": 13, "name": "spleeter (vocals, drums, bass, piano, other)", "algorithm_group_id": 1, "orientation": 0, "render_id": 2, "order_id": 13, "created_at": null, "updated_at": null, "additional_fields": null, "price_coefficient": 1, "is_active": 1, "algorithm_fields": [], "algorithm_descriptions": [] }, { "id": 14, "name": "unmix XL (vocals, drums, bass, other)", "algorithm_group_id": 1, "orientation": 0, "render_id": 3, "order_id": 14, "created_at": null, "updated_at": null, "additional_fields": null, "price_coefficient": 1, "is_active": 1, "algorithm_fields": [], "algorithm_descriptions": [] }, { "id": 15, "name": "unmix HQ (vocals, drums, bass, other)", "algorithm_group_id": 1, "orientation": 0, "render_id": 4, "order_id": 15, "created_at": null, "updated_at": null, "additional_fields": null, "price_coefficient": 1, "is_active": 1, "algorithm_fields": [], "algorithm_descriptions": [] }, { "id": 16, "name": "unmix SD (vocals, drums, bass, other)", "algorithm_group_id": 1, "orientation": 0, "render_id": 5, "order_id": 16, "created_at": null, "updated_at": null, "additional_fields": null, "price_coefficient": 1, "is_active": 1, "algorithm_fields": [], "algorithm_descriptions": [] }, { "id": 17, "name": "unmix SE (vocals, music)", "algorithm_group_id": 1, "orientation": 0, "render_id": 6, "order_id": 17, "created_at": null, "updated_at": null, "additional_fields": null, "price_coefficient": 1, "is_active": 1, "algorithm_fields": [], "algorithm_descriptions": [] }, { "id": 18, "name": "Zero Shot (Query Based) (Low quality)", "algorithm_group_id": 1, "orientation": 0, "render_id": 14, "order_id": 18, "created_at": null, "updated_at": null, "additional_fields": null, "price_coefficient": 1, "is_active": 1, "algorithm_fields": [ { "id": 6, "name": "add_opt1", "text": "Model Type", "algorithm_id": 18, "options": "{\n \"bass_0\": \"Bass (MUSDB18HQ AVG)\",\n \"drums_1\": \"Drums (MUSDB18HQ AVG)\",\n \"vocals_2\": \"Vocals (MUSDB18HQ AVG)\",\n \"other_3\": \"Other (MUSDB18HQ AVG)\"\n}", "created_at": "2022-12-07 11:10:14", "updated_at": "2023-03-12 02:07:23", "default_key": null, "server_key": "vector" } ], "algorithm_descriptions": [] }, { "id": 19, "name": "MDX A (vocals, drums, bass, other)", "algorithm_group_id": 1, "orientation": 0, "render_id": 7, "order_id": 19, "created_at": null, "updated_at": null, "additional_fields": null, "price_coefficient": 1, "is_active": 1, "algorithm_fields": [], "algorithm_descriptions": [] }, { "id": 20, "name": "Demucs3 Model A (vocals, drums, bass, other)", "algorithm_group_id": 1, "orientation": 0, "render_id": 10, "order_id": 20, "created_at": null, "updated_at": null, "additional_fields": null, "price_coefficient": 1, "is_active": 1, "algorithm_fields": [], "algorithm_descriptions": [ { "id": 5, "short_description": "Algorithm Demucs3 (Simple)", "long_description": "

Algorithm Demucs3 splits track into 4 stems (bass, drums, vocals, other). The winner of the Music Demuxing Challenge 2021. Only MUSDB18 training data was used for training of model, so quality is worse than Demucs3 Model B. Demucs3 Model A and Demucs3 Model B has the same architecture, but has different weights.

\r\n

Quality table

\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n
Multisong datasetSynth dataset
SDR BassSDR DrumsSDR OtherSDR VocalsSDR InstrumentalSDR VocalsSDR Instrumental
9.508.974.407.2113.52------
", "lang": "en", "algorithm_id": 20, "created_at": "2023-06-12 02:23:57", "updated_at": "2023-06-12 13:38:16" } ] } ] } } } } } }, "/api/app/register": { "post": { "summary": "Register", "responses": { "200": { "description": "Register", "headers": { "Date": { "schema": { "type": "string" }, "example": "Mon, 05 Feb 2024 14:42:26 GMT" }, "Server": { "schema": { "type": "string" }, "example": "Apache/2.4.52 (Ubuntu)" }, "Cache-Control": { "schema": { "type": "string" }, "example": "no-cache, private" }, "X-RateLimit-Limit": { "schema": { "type": "integer" }, "example": "60" }, "X-RateLimit-Remaining": { "schema": { "type": "integer" }, "example": "59" }, "Keep-Alive": { "schema": { "type": "string" }, "example": "timeout=5, max=100" }, "Connection": { "schema": { "type": "string" }, "example": "Keep-Alive" }, "Transfer-Encoding": { "schema": { "type": "string" }, "example": "chunked" } }, "content": { "application/json": { "schema": { "type": "object", "properties": { "success": { "type": "boolean" }, "message": { "type": "string" } } }, "example": { "success": true, "message": "Account verification email sent to asdf@gmail.com" } } } } } } }, "/api/app/login": { "post": { "summary": "Login", "responses": { "200": { "description": "Login", "headers": { "Server": { "schema": { "type": "string" }, "example": "nginx" }, "Date": { "schema": { "type": "string" }, "example": "Wed, 22 May 2024 12:25:06 GMT" }, "Transfer-Encoding": { "schema": { "type": "string" }, "example": "chunked" }, "Connection": { "schema": { "type": "string" }, "example": "keep-alive" }, "Vary": { "schema": { "type": "string" }, "example": "Accept-Encoding" }, "Cache-Control": { "schema": { "type": "string" }, "example": "private, must-revalidate" }, "pragma": { "schema": { "type": "string" }, "example": "no-cache" }, "expires": { "schema": { "type": "integer" }, "example": "-1" }, "X-RateLimit-Limit": { "schema": { "type": "integer" }, "example": "120" }, "X-RateLimit-Remaining": { "schema": { "type": "integer" }, "example": "118" }, "Content-Encoding": { "schema": { "type": "string" }, "example": "gzip" } }, "content": { "application/json": { "schema": { "type": "object", "properties": { "success": { "type": "boolean" }, "data": { "type": "object", "properties": { "name": { "type": "string" }, "email": { "type": "string" }, "api_token": { "nullable": true }, "created_at": { "type": "string", "format": "style" }, "updated_at": { "type": "string" }, "premium_minutes": { "type": "integer" }, "premium_enabled": { "type": "integer" }, "long_filenames_enabled": { "type": "integer" } } } } }, "example": { "success": true, "data": { "name": "", "email": "", "api_token": null, "created_at": "2023-09-29 15:05:50", "updated_at": "2024-01-24 16:28:33", "premium_minutes": 54, "premium_enabled": 1, "long_filenames_enabled": 1 } } } } }, "400": { "description": "Login", "headers": { "Date": { "schema": { "type": "string" }, "example": "Wed, 22 May 2024 10:39:49 GMT" }, "Server": { "schema": { "type": "string" }, "example": "Apache/2.4.52 (Ubuntu)" }, "Cache-Control": { "schema": { "type": "string" }, "example": "no-cache, private" }, "X-RateLimit-Limit": { "schema": { "type": "integer" }, "example": "120" }, "X-RateLimit-Remaining": { "schema": { "type": "integer" }, "example": "118" }, "Connection": { "schema": { "type": "string" }, "example": "close" }, "Transfer-Encoding": { "schema": { "type": "string" }, "example": "chunked" } }, "content": { "application/json": { "schema": { "type": "object", "properties": { "success": { "type": "boolean" }, "message": { "type": "object", "properties": { "email": { "type": "array", "items": { "type": "string" } }, "password": { "type": "array", "items": { "type": "string" } } } } } }, "example": { "success": false, "message": { "email": [ "The email field is required." ], "password": [ "The password field is required." ] } } } } } } } }, "/api/separation/create": { "post": { "summary": "Separation", "responses": {} } }, "/api/separation/get": { "get": { "summary": "Separation", "parameters": [ { "name": "hash", "in": "query", "schema": { "type": "string" }, "example": "20240119124024-413c481b80-10-luke-s-theme-a-madman-s-symphony.mp3" } ], "responses": { "200": { "description": "Separation", "headers": { "Date": { "schema": { "type": "string" }, "example": "Mon, 05 Feb 2024 14:49:20 GMT" }, "Server": { "schema": { "type": "string" }, "example": "Apache/2.4.52 (Ubuntu)" }, "Cache-Control": { "schema": { "type": "string" }, "example": "no-cache, private" }, "X-RateLimit-Limit": { "schema": { "type": "integer" }, "example": "60" }, "X-RateLimit-Remaining": { "schema": { "type": "integer" }, "example": "58" }, "Keep-Alive": { "schema": { "type": "string" }, "example": "timeout=5, max=100" }, "Connection": { "schema": { "type": "string" }, "example": "Keep-Alive" }, "Transfer-Encoding": { "schema": { "type": "string" }, "example": "chunked" } }, "content": { "application/json": { "schema": { "type": "object", "properties": { "success": { "type": "boolean" }, "status": { "type": "string" }, "data": { "type": "object", "properties": { "queue_count": { "type": "integer" }, "current_order": { "type": "integer" } } } } }, "example": { "success": true, "status": "waiting", "data": { "queue_count": 3, "current_order": 2 } } } } } } } }, "/api/app/news": { "get": { "summary": "News", "parameters": [ { "name": "lang", "in": "query", "schema": { "type": "string" }, "example": "ru" }, { "name": "start", "in": "query", "schema": { "type": "integer" }, "example": "2" }, { "name": "limit", "in": "query", "schema": { "type": "integer" }, "example": "1" } ], "responses": { "200": { "description": "News", "headers": { "Date": { "schema": { "type": "string" }, "example": "Fri, 10 May 2024 08:06:07 GMT" }, "Server": { "schema": { "type": "string" }, "example": "Apache/2.4.52 (Ubuntu)" }, "Cache-Control": { "schema": { "type": "string" }, "example": "no-cache, private" }, "X-RateLimit-Limit": { "schema": { "type": "integer" }, "example": "120" }, "X-RateLimit-Remaining": { "schema": { "type": "integer" }, "example": "119" }, "Keep-Alive": { "schema": { "type": "string" }, "example": "timeout=5, max=100" }, "Connection": { "schema": { "type": "string" }, "example": "Keep-Alive" }, "Transfer-Encoding": { "schema": { "type": "string" }, "example": "chunked" } }, "content": { "application/json": { "schema": { "type": "array", "items": { "type": "object", "properties": { "id": { "type": "integer" }, "title": { "type": "string" }, "lang": { "type": "string" }, "text": { "type": "string", "format": "style" }, "status": { "type": "integer" }, "created_at": { "type": "string" }, "updated_at": { "type": "string", "format": "style" } } } }, "example": [ { "id": 33, "title": "Обновления в декабре", "lang": "ru", "text": "

1) Мы добавили новую модель BandIt Plus для разделения треков на голос, музыку и эффекты. Она может быть полезна для телевизионных или кинематографических роликов. Модель была выложена авторами статьи \"A Generalized Bandsplit Neural Network for Cinematic Audio Source Separation\" в репозитории на GitHub. Модель была натренирована на наборе данных Divide and Remaster (DnR). И на данный момент имеет лучшие метрики качества среди подобных моделей. Демо работы алгоритма можно посмотреть здесь.

\r\n

Таблица качества

\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n\r\n
Название алгоритмаDnR dataset
SDR SpeechSDR MusicSDR Effects
BandIt Plus15.629.219.69
\r\n

2) Код почти для всех моделей был обновлен таким образом, что качество разделения немного выросло, а скорость разделения увеличилась.

\r\n

3) Модель для удаления звуков толпы (Crowd removal) была обновлена. Добавлено более качественное удаление смеха.

", "status": 1, "created_at": "2023-12-21 05:00:00", "updated_at": "2023-12-22 12:15:36" } ] } } } } } }, "/api/app/demo": { "get": { "summary": "Demo", "responses": { "200": { "description": "Demo", "headers": { "Server": { "schema": { "type": "string" }, "example": "nginx" }, "Date": { "schema": { "type": "string" }, "example": "Tue, 14 May 2024 12:37:19 GMT" }, "Transfer-Encoding": { "schema": { "type": "string" }, "example": "chunked" }, "Connection": { "schema": { "type": "string" }, "example": "keep-alive" }, "Vary": { "schema": { "type": "string" }, "example": "Accept-Encoding" }, "Cache-Control": { "schema": { "type": "string" }, "example": "private, must-revalidate" }, "pragma": { "schema": { "type": "string" }, "example": "no-cache" }, "expires": { "schema": { "type": "integer" }, "example": "-1" }, "X-RateLimit-Limit": { "schema": { "type": "integer" }, "example": "120" }, "X-RateLimit-Remaining": { "schema": { "type": "integer" }, "example": "118" }, "Content-Encoding": { "schema": { "type": "string" }, "example": "gzip" } }, "content": { "application/json": { "schema": { "type": "array", "items": { "type": "object", "properties": { "hash": { "type": "string" }, "date": { "type": "string", "format": "style" }, "input_audio": { "type": "string", "format": "uri" }, "algorithm": { "type": "object", "properties": { "name": { "type": "string" }, "algorithm_group_id": { "type": "integer" }, "orientation": { "type": "integer" }, "render_id": { "type": "integer" }, "order_id": { "type": "integer" }, "created_at": { "type": "string", "format": "style" }, "updated_at": { "type": "string" } } } } } }, "example": [ { "hash": "20240514123536-15f8f16001-lxst-cxntury-carbon-mist.mp3", "date": "2024-05-14 12:35", "input_audio": "https://mvsep.com/storage/temp/20240514123536-15f8f16001-lxst-cxntury-carbon-mist.mp3", "algorithm": { "name": "DrumSep (kick, snare, cymbals, toms)", "algorithm_group_id": 2, "orientation": 0, "render_id": 37, "order_id": 400, "created_at": "2023-12-24 11:43:42", "updated_at": "2023-12-24 11:45:03" } }, { "hash": "20240514122600-e410ae9535-justin-bieber-sorry-lyrics-128k.m4a", "date": "2024-05-14 12:26", "input_audio": "https://mvsep.com/storage/temp/20240514122600-e410ae9535-justin-bieber-sorry-lyrics-128k.m4a", "algorithm": { "name": "MDX-B Karaoke (lead/back vocals)", "algorithm_group_id": 2, "orientation": 1, "render_id": 12, "order_id": 70, "created_at": null, "updated_at": "2024-03-08 08:19:07" } } ] } } } } } } } }