Sawtak Arabi documentation
Generate Arabic speech for your app or voice agent. Choose a voice, connect to the API, and work with streamed audio.
Generate Arabic speech for your voice agent, assistant or application. Choose a voice and send text using the API directly or through the OpenAI SDK.
- Quickstart — generate your first audio file.
- Use with an AI agent — generate voiceovers with Claude Code or Codex.
- Arabic TTS integrations — choose between the API, hosted MCP and coding-agent skill.
- Hosted MCP — connect your assistant with OAuth.
- Authentication — create and configure an API key.
API endpoints
Base URL: https://api.sawtakarabi.ai/v1. Send your API key with each request. Public voice previews work without a key.
| Endpoint | Purpose |
|---|---|
POST /audio/speech | Generate speech. |
GET /voices, GET /voices/{id} | Find and inspect voices. |
GET /voices/{id}/preview | Play a voice sample. |
POST /voices, PATCH /voices/{id}, DELETE /voices/{id} | Clone and manage voices. |
GET /operations/{id}, POST /operations/{id}/cancel | Check or cancel a request. |
GET /me, GET /usage | Check your balance and usage. |
GET /history, GET /history/{id}, GET /history/{id}/audio | Retrieve saved recordings. |
GET /models, GET /models/arabic-tts-1 | Discover the arabic-tts-1 model. |
See Streaming audio, Request limits, and Errors when building your integration.
Machine-readable documentation: llms.txt and all guides.