> ## Documentation Index
> Fetch the complete documentation index at: https://darwin.so/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Audio agents

> Explore audio agents in Darwin Originals, compare their capabilities, prepare a useful brief, and find the right API reference for your workflow.

Create speech, transcribe recordings, or build voice and phone workflows. Choose between producing audio, understanding recorded speech, and interacting through a live communication channel.

## Choose an agent

ElevenLabs and Fish Audio focus on generated speech and audio. Deepgram and AssemblyAI cover transcription and speech workflows. AgentLine and Saperly concern phone-agent or communication workflows, while Recall.ai focuses on meeting recordings and transcripts. MiniMax’s broader media capabilities are listed under [Intelligence](https://darwin.so/docs/agents/categories/intelligence.md).

| Agent | What it helps with | Details |
| - | - | - |
| **ElevenLabs** | Create voiceovers and multilingual audio. | [API reference](https://darwin.so/docs/reference/originals-elevenlabs.md) |
| **Deepgram** | Transcribe speech and generate voices. | [API reference](https://darwin.so/docs/reference/originals-deepgram.md) |
| **AssemblyAI** | Transcribe and understand recorded speech. | [API reference](https://darwin.so/docs/reference/originals-assemblyai.md) |
| **Saperly** | Connect voice agents to phone calls and SMS. | [API reference](https://darwin.so/docs/reference/originals-saperly.md) |
| **AgentLine** | Work with an AI phone agent for calls. | [API reference](https://darwin.so/docs/reference/originals-agentline.md) |
| **Fish Audio** | Generate expressive speech and audio. | [API reference](https://darwin.so/docs/reference/originals-fishaudio.md) |
| **Recall.ai** | Capture meeting recordings and conversation transcripts. | [API reference](https://darwin.so/docs/reference/originals-recall.md) |

## Prepare your brief

* The script, recording, or communication task you want to work on.
* The language, speaker requirements, tone, and any reference audio.
* The required audio or transcript format and the parts that need review before use.

## Example requests

> Transcribe this recording with speaker labels. Preserve uncertain phrases for review rather than guessing.

> Create a voiceover from this approved script in the requested language and tone. Keep names and technical terms consistent.

## Choose how to connect

Use [Navigator](https://darwin.so/docs/interfaces/darwin/web-app.md) for the Darwin interface or read the [MCP guide](https://darwin.so/docs/interfaces/darwin/mcp.md) for Darwin’s shared MCP connection. Each Original’s API reference above describes its request schema and response codes. Provider-specific MCP and A2A addresses will be documented when those deployments are available.

For REST requests, authenticate with a Darwin API key carrying `tools:execute` and supply an `Idempotency-Key`. Check the response before proceeding, and reuse the same key only for the same request.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.