Speechbase

Stream speech

Streams audio from the provider for low latency (provider pass-through). Pass either `voiceId` (to use a saved Voice) or `model` + `voice` (inline), not both. Whole-clip params (`volumeDbfs`, `output` format conversion) are not accepted here. Use POST /v1/audio/speech for those.

POST
/v1/audio/speech/stream

Streams audio from the provider for low latency (provider pass-through). Pass either voiceId (to use a saved Voice) or model + voice (inline), not both. Whole-clip params (volumeDbfs, output format conversion) are not accepted here. Use POST /v1/audio/speech for those.

Authorization

bearerAuth
AuthorizationBearer <token>

API key

In: header

Request Body

application/json

TypeScript Definitions

Use the request body type in TypeScript.

Response Body

application/problem+json

application/problem+json

application/json

curl -X POST "https://example.com/v1/audio/speech/stream" \  -H "Authorization: Bearer $SPEECHBASE_API_KEY" \  -H "Content-Type: application/json" \  -d '{  "voiceId": "550e8400-e29b-41d4-a716-446655440000",  "text": "Hello from a saved voice."}'
"string"
{
  "type": "string",
  "title": "string",
  "status": 0,
  "detail": "string",
  "code": "string",
  "validation": [
    {
      "path": [
        "string"
      ],
      "message": "string"
    }
  ],
  "provider": "string",
  "upstream_code": "string",
  "upstream_status": 0,
  "turn_index": 0
}
{
  "type": "string",
  "title": "string",
  "status": 0,
  "detail": "string",
  "code": "string",
  "validation": [
    {
      "path": [
        "string"
      ],
      "message": "string"
    }
  ],
  "provider": "string",
  "upstream_code": "string",
  "upstream_status": 0,
  "turn_index": 0
}
{
  "error": {
    "code": "content_moderation_blocked",
    "message": "string",
    "reason": {
      "type": "error_fail_closed"
    }
  }
}

Generate speech POST

Synthesizes the whole clip and returns raw audio bytes in a single response. Pass either `voiceId` (to use a saved Voice) or `model` + `voice` (inline), not both. Because the full clip is produced server-side, whole-clip params (`volumeDbfs`, `output` format conversion) are applied here. For low-latency provider streaming, use POST /v1/audio/speech/stream.

Generate speech with timestamps POST

Synthesizes speech and returns a JSON envelope with base64 audio and word-level timestamps. Pass either `voiceId` (to use a saved Voice) or `model` + `voice` (inline) — not both.