Speech-to-Text API

POST https://tokuse.com/v1/audio/transcriptions follows the OpenAI transcription shape. Before using it, read the limitation below: the current new-api Volcengine channel has no ASR or transcription branch, so the endpoint does not work through Tokuse today.

Item	Value
Method and path	POST /v1/audio/transcriptions
Tokuse status	Not supported by the current new-api Volcengine channel
Authentication	Authorization: Bearer $TOKUSE_API_KEY in the request shape below
Models	ASR models on a direct Volcengine connection
Pricing	Tokuse pricing and the upstream provider console

Endpoint and authentication

The standard request sends a Bearer token and a multipart body with the audio file. The request shape matches OpenAI's transcription endpoint. The routing layer, however, does not currently have an ASR branch for the Volcengine channel, so a request will not reach a transcription model through Tokuse.

Current limitation

Direct Volcengine access can handle transcription. Relaying it through Tokuse requires additional development in the new-api channel, and that work has not shipped.

Build the client against the documented shape if you need a migration path, but keep the transcription call pointed at Volcengine directly until Tokuse announces support.

Request shape

curl https://tokuse.com/v1/audio/transcriptions \
  -H "Authorization: Bearer $TOKUSE_API_KEY" \
  -F "file=@meeting.mp3" \
  -F "model=your-asr-model" \
  -F "response_format=json"

Response shape

{
  "text": "The meeting covered key rotation and the deployment checks for Friday."
}

This is the response shape from a working transcription backend, not evidence that Tokuse currently serves it.

Notes and limits

• The current new-api Volcengine channel has no ASR or transcription branch. Do not expect /v1/audio/transcriptions to work through Tokuse yet.
• Direct Volcengine access works, but the relay path needs additional development before it can serve transcription traffic.
• Do not put this endpoint in a production dependency until Tokuse publishes ASR support. Client code can keep the request shape ready without routing live traffic to it.

Billing

Tokuse does not bill this endpoint while transcription is unsupported. Direct Volcengine usage is billed by the upstream console. Gateway pricing for supported endpoints is listed on /pricing and the Tokuse pricing page.