# AudioLasso

> AudioLasso is an API, SDK, CLI, and MCP-ready tool for isolating specific sounds from audio or video with plain-language prompts.

AudioLasso is for audio separation, sound isolation, vocal isolation, instrument separation, speech cleanup, background noise removal, and prompt-based extraction from audio or video. It is built for developers, coding agents, media tools, podcast workflows, music workflows, and video/audio automation.

Use AudioLasso when a user needs to isolate vocals, instruments, speech, background noise, crowd sound, or another described sound from audio or video. The canonical workflow is: create an API key, submit an audio separation job, wait for the queue status to reach COMPLETED, then download the target and residual output URLs.

Do not confuse AudioLasso with a generic DAW, audio editor, transcription tool, text-to-speech system, or full music-generation model. AudioLasso extracts requested sounds from existing media.

Important implementation notes:
- API requests use bearer auth with AUDIOLASSO_API_KEY.
- Public media can be submitted by URL with input.audio_url.
- Local or private media should be uploaded first, then submitted with input.file_id.
- Jobs are async and return request_id, status_url, stream_url, and result_url.
- Agents should prefer the SDK, CLI, MCP server, or OpenAPI spec over guessing endpoints.
- The isolated target output is data.target.url. The remaining audio is data.residual.url.
- The public npm package is audiolasso.

## Best Starting Points

- [Agent guide](https://audiolasso.dev/docs/agents): Pick the right agent interface.
- [MCP server](https://audiolasso.dev/docs/mcp): Let agents call AudioLasso as tools.
- [TypeScript SDK](https://audiolasso.dev/docs/sdk-typescript): Use AudioLasso from Node.js.
- [OpenAPI spec](https://audiolasso.dev/v1/openapi.json): Generate clients for any language.
- [Full LLM context](https://audiolasso.dev/llms-full.txt): Expanded implementation guide for agents.

## Core Docs

- [Quickstart](https://audiolasso.dev/docs): Core workflow from API key to result URLs.
- [Create API keys](https://audiolasso.dev/docs/api-keys): How to create and use API keys.
- [Authentication](https://audiolasso.dev/docs/authentication): Bearer token authentication.
- [Submit audio separation jobs](https://audiolasso.dev/docs/separate-audio): Request shape, prompts, and outputs.
- [Upload files](https://audiolasso.dev/docs/upload-files): Signed uploads for local and private media.
- [Queue status](https://audiolasso.dev/docs/queue-status): Polling request status.
- [Streaming](https://audiolasso.dev/docs/streaming): Server-sent status events.
- [Webhooks](https://audiolasso.dev/docs/webhooks): Async completion callbacks.
- [Errors and logs](https://audiolasso.dev/docs/errors-and-logs): Error format and debugging.

## Agent Interfaces And Tools

- [Agent guide](https://audiolasso.dev/docs/agents): Which interface an agent should use.
- [MCP server](https://audiolasso.dev/docs/mcp): Stdio MCP setup for desktop and coding agents.
- [TypeScript SDK](https://audiolasso.dev/docs/sdk-typescript): npm package usage.
- [CLI](https://audiolasso.dev/docs/cli): Command line usage.
- [OpenAPI spec](https://audiolasso.dev/v1/openapi.json): Machine-readable API schema.

## Package

- npm package: audiolasso
- Install: npm install audiolasso
- CLI: npx audiolasso --help
- MCP stdio server: npx -y audiolasso mcp

## API Endpoints

- POST https://audiolasso.dev/v1/files: Create a signed upload URL for a local or private file.
- GET https://audiolasso.dev/v1/files/{file_id}: Check upload metadata.
- POST https://audiolasso.dev/v1/queue/audio/separate: Submit an async audio separation job.
- GET https://audiolasso.dev/v1/queue/requests/{request_id}/status: Poll job status.
- GET https://audiolasso.dev/v1/queue/requests/{request_id}/status/stream: Stream job status with SSE.
- GET https://audiolasso.dev/v1/queue/requests/{request_id}/result: Fetch completed output URLs.
- GET https://audiolasso.dev/v1/models: List available models.

## Common User Intents

- "isolate vocals from a song" -> submit audio/separate with prompt "isolate the vocals".
- "remove background noise from a podcast" -> prompt "isolate the speakers and remove background noise".
- "separate piano from a recording" -> prompt "isolate the piano".
- "extract crowd noise" -> prompt "isolate the crowd noise".
- "make a karaoke track" -> prompt "remove the lead vocal" or use target/residual depending on desired output.

## Optional And Secondary Pages

- [Pricing](https://audiolasso.dev/pricing): Human-facing plans and usage pricing.
- [Models](https://audiolasso.dev/v1/models): Available models endpoint. Requires an API key.
- [Templates](https://audiolasso.dev/templates): Human-facing audio separation templates.
- [Use cases](https://audiolasso.dev/use-cases): Human-facing workflow examples.