> ## Documentation Index > Fetch the complete documentation index at: https://docs.eachlabs.ai/llms.txt > Use this file to discover all available pages before exploring further. # Stream synthesized speech > OpenAI-compatible text-to-speech. Use `Authorization: Bearer ` and a JSON body. The response is a raw audio byte stream forwarded progressively (no prediction polling); the negotiated content type is returned verbatim (`audio/mpeg` for mp3, `audio/pcm;rate=…;channels=…` for pcm). An omitted `response_format` defaults to `mp3`. The accepted model/voice/format set is a curated per-environment catalog that a model joins only after passing contract qualification; `deepgram/aura-2` is the launch model. An unsupported model, voice, or format, or a `speed`/`instructions` value for a model that does not support it, returns a 400 without dispatching a synthesis. This endpoint is independently feature-gated during preview rollout. A disabled deployment returns an OpenAI-shaped 404 without dispatching. ## OpenAPI ````yaml /openapi_specs/models.json post /v1/audio/speech openapi: 3.0.3 info: title: each::labs Official API v1 description: > Official each::labs API V1. This API provides access to our comprehensive AI model catalog and other resources. ## Public contract notes - Public API routes handled by request logging include `X-Request-Id` for support and log correlation. Clients may send `X-Request-Id`; otherwise api-service generates one. Root, health, and metrics endpoints are excluded. - REST error bodies keep the backward-compatible JSON shape used by existing clients. Most errors include at least `error`; some shared error paths also include `status` and `details`. Use the HTTP status code as the primary programmatic signal unless an endpoint documents a richer envelope. - Rate-limit and retry headers are endpoint-dependent. If a response includes `Retry-After`, honor it before retrying; otherwise use bounded exponential backoff for retryable 429/5xx responses. version: 1.0.0 contact: name: API Support email: support@eachlabs.ai x-logo: url: ./logo-white.svg altText: each::labs servers: - url: https://api.eachlabs.ai description: Production server security: [] tags: - name: AI Models description: Endpoints for listing and retrieving AI models and other resources. - name: AI Models Prediction description: Endpoints for managing model predictions - name: Uploads description: Endpoints for uploading files that can be used as model inputs. - name: Executions description: Endpoints for listing execution history. - name: Workflows description: >- Endpoints for creating workflows and triggering workflow executions through api-service. - name: OpenAI Compatibility description: OpenAI-compatible endpoints for LLM Router requests. - name: Webhooks description: > Endpoints for retrieving webhook information and delivery history. **Note:** Currently, webhooks are only supported for Workflows V2. We are actively expanding webhook support to other services and will update this documentation as new integrations become available. paths: /v1/audio/speech: post: tags: - OpenAI Compatibility summary: Stream synthesized speech description: | OpenAI-compatible text-to-speech. Use `Authorization: Bearer ` and a JSON body. The response is a raw audio byte stream forwarded progressively (no prediction polling); the negotiated content type is returned verbatim (`audio/mpeg` for mp3, `audio/pcm;rate=…;channels=…` for pcm). An omitted `response_format` defaults to `mp3`. The accepted model/voice/format set is a curated per-environment catalog that a model joins only after passing contract qualification; `deepgram/aura-2` is the launch model. An unsupported model, voice, or format, or a `speed`/`instructions` value for a model that does not support it, returns a 400 without dispatching a synthesis. This endpoint is independently feature-gated during preview rollout. A disabled deployment returns an OpenAI-shaped 404 without dispatching. operationId: createAudioSpeech requestBody: required: true content: application/json: schema: $ref: '#/components/schemas/AudioSpeechRequest' responses: '200': description: | Raw audio byte stream. Bytes are forwarded progressively as the provider synthesizes them; the content type is the negotiated format (`audio/mpeg` or `audio/pcm;rate=…;channels=…`). headers: X-Generation-Id: description: >- Upstream generation identifier for out-of-band billing reconciliation. schema: type: string X-Execution-ID: description: Eachlabs execution identifier for the synthesis. schema: type: string content: audio/mpeg: schema: type: string format: binary audio/pcm: schema: type: string format: binary '400': description: Unsupported model/voice/format/param or invalid input content: application/json: schema: $ref: '#/components/schemas/OpenAIErrorResponse' '401': description: Authentication failed content: application/json: schema: $ref: '#/components/schemas/OpenAIErrorResponse' '402': description: Insufficient balance content: application/json: schema: $ref: '#/components/schemas/OpenAIErrorResponse' '404': description: Capability disabled for this deployment content: application/json: schema: $ref: '#/components/schemas/OpenAIErrorResponse' '413': description: Request body exceeds the maximum size content: application/json: schema: $ref: '#/components/schemas/OpenAIErrorResponse' '429': description: Provider rate limit reached content: application/json: schema: $ref: '#/components/schemas/OpenAIErrorResponse' '500': description: Server error content: application/json: schema: $ref: '#/components/schemas/OpenAIErrorResponse' '503': description: Authentication or speech service temporarily unavailable content: application/json: schema: $ref: '#/components/schemas/OpenAIErrorResponse' '504': description: Speech request timed out content: application/json: schema: $ref: '#/components/schemas/OpenAIErrorResponse' security: - BearerAuth: [] components: schemas: AudioSpeechRequest: type: object required: - model - input - voice properties: model: type: string example: deepgram/aura-2 description: >- Must be in the environment's speech catalog. An unlisted model returns 400 without dispatching a synthesis. input: type: string description: >- Text to synthesize. Bounded by the model's per-character limit (counted in Unicode runes). voice: type: string example: aura-2-thalia-en description: >- Model-specific, namespaced voice id. Must be allowlisted for the model. response_format: type: string default: mp3 description: >- Audio format. Omitted defaults to `mp3`. Allowed values are per-model (e.g. `mp3`, `pcm`). example: mp3 speed: type: number minimum: 0.25 maximum: 4 description: >- Playback speed. Accepted only for models that declare support; otherwise 400. instructions: type: string description: >- Steering instructions. Accepted only for models that declare support; otherwise 400. stream_format: type: string enum: - audio description: Only raw audio streaming is supported. `sse` is rejected with 400. OpenAIErrorResponse: type: object description: >- OpenAI-compatible error envelope used by chat, audio, and realtime compatibility endpoints. properties: error: type: object properties: message: type: string example: Invalid API key type: type: string example: authentication_error param: type: string example: model code: type: string example: invalid_api_key securitySchemes: BearerAuth: type: http scheme: bearer description: >- API key passed as an Authorization Bearer token: `Authorization: Bearer YOUR_API_KEY` ````