Speech-to-Text (STT)

Speech-to-Text transcribes audio files into text. Otoroshi exposes STT through an OpenAI-compatible API endpoint.

Supported providers and models

Provider	Models
OpenAI	whisper-1, gpt-4o-mini-transcribe
Azure OpenAI	whisper-1, gpt-4o-mini-transcribe
Cloud Temple 🇫🇷 🇪🇺	whisper-1, gpt-4o-mini-transcribe
Groq	whisper-large-v3, whisper-large-v3-turbo, distil-whisper-large-v3-en
ElevenLabs	scribe_v1
Mistral 🇫🇷 🇪🇺	voxtral-mini-latest, voxtral-mini-2507
AlphaEdge 🇫🇷 🇪🇺	alpha-audio-v1

STT configuration

OpenAI / Azure OpenAI / Cloud Temple

{
  "stt": {
    "enabled": true,
    "model": "whisper-1",
    "language": "en",
    "prompt": "Optional context for better transcription",
    "response_format": "json",
    "temperature": 0
  }
}

Parameter	Type	Default	Description
enabled	boolean	true	Enable or disable STT
model	string	—	The STT model to use
language	string	—	Language of the audio (ISO 639-1 code, e.g. en, fr, de)
prompt	string	—	Optional text to guide the model's transcription style
response_format	string	—	Response format: json, text, srt, verbose_json, vtt
temperature	number	—	Sampling temperature between 0 and 1

Groq

Same parameters as OpenAI.

{
  "stt": {
    "enabled": true,
    "model": "whisper-large-v3-turbo",
    "language": "en"
  }
}

ElevenLabs

{
  "stt": {
    "enabled": true,
    "model_id": "scribe_v1",
    "language": "en"
  }
}

Parameter	Type	Default	Description
enabled	boolean	true	Enable or disable STT
model_id	string	scribe_v1	The ElevenLabs STT model
language	string	—	Language code for transcription

Mistral (Voxtral)

{
  "stt": {
    "enabled": true,
    "model": "voxtral-mini-latest",
    "language": "fr",
    "diarize": true,
    "temperature": 0
  }
}

Parameter	Type	Default	Description
enabled	boolean	true	Enable or disable STT
model	string	—	The Voxtral model to use
language	string	—	Language of the audio
diarize	boolean	—	Enable speaker diarization (identifies different speakers in the audio). Mistral-specific feature.
temperature	number	—	Sampling temperature

AlphaEdge

AlphaEdge 🇫🇷 🇪🇺 is a French/EU provider specialized in speech transcription and OCR. The audio model only supports STT (no TTS, no translation). Authentication uses the X-API-Key header (set through config.connection.token).

{
  "stt": {
    "enabled": true,
    "model": "alpha-audio-v1",
    "enable_diarization": false,
    "enable_postcorrect": false
  }
}

Parameter	Type	Default	Description
enabled	boolean	true	Enable or disable STT
model	string	alpha-audio-v1	The AlphaEdge transcription model
enable_diarization	boolean	false	Enable speaker diarization (identifies different speakers in the audio)
enable_postcorrect	boolean	false	Apply linguistic post-correction (punctuation, capitalization, spelling, stuttering removal)

enable_diarization and enable_postcorrect can also be passed per-request as form fields, overriding the entity configuration.

API usage

Plugin setup

Add the Cloud APIM - Speech to text backend plugin to your route:

{
  "enabled": true,
  "plugin": "cp:otoroshi_plugins.com.cloud.apim.otoroshi.extensions.aigateway.plugins.OpenAICompatSpeechToText",
  "config": {
    "refs": ["audio-gen-model_xxxxxxxxx"],
    "max_size_upload": 104857600
  }
}

Parameter	Type	Default	Description
refs	array of strings	—	References to audio model entities
max_size_upload	number	104857600 (100 MB)	Maximum upload file size in bytes

Request

curl https://my-audio-endpoint.example.com/v1/audio/transcriptions \
  -H "Authorization: Bearer $OTOROSHI_API_KEY" \
  -F "file=@recording.mp3" \
  -F "model=whisper-1" \
  -F "language=en"

The request is sent as multipart form data.

Request parameters

Parameter	Type	Description
file	file	The audio file to transcribe (required)
model	string	Model to use. Supports provider_id###model_name routing syntax
language	string	Language of the audio
prompt	string	Optional context for better transcription
response_format	string	Desired response format
temperature	string	Sampling temperature

Response

{
  "text": "Hello, how are you today?"
}

Full entity example with Mistral (Voxtral)

{
  "id": "audio-gen-model_xxxxxxxxx",
  "name": "Mistral Voxtral STT",
  "description": "Mistral speech-to-text with Voxtral",
  "provider": "mistral",
  "config": {
    "connection": {
      "token": "${vault://local/MISTRAL_API_TOKEN}",
      "timeout": 30000
    },
    "options": {
      "stt": {
        "enabled": true,
        "model": "voxtral-mini-latest",
        "diarize": true
      }
    }
  },
  "kind": "ai-gateway.extensions.cloud-apim.com/AudioModel"
}

Full entity example with ElevenLabs

{
  "id": "audio-gen-model_xxxxxxxxx",
  "name": "ElevenLabs Audio",
  "description": "ElevenLabs TTS and STT",
  "provider": "elevenlabs",
  "config": {
    "connection": {
      "token": "${vault://local/ELEVENLABS_API_KEY}",
      "timeout": 30000
    },
    "options": {
      "tts": {
        "enabled": true,
        "model_id": "eleven_multilingual_v2",
        "voice_id": "21m00Tcm4TlvDq8ikWAM",
        "output_format": "mp3_44100_128"
      },
      "stt": {
        "enabled": true,
        "model_id": "scribe_v1"
      }
    }
  },
  "kind": "ai-gateway.extensions.cloud-apim.com/AudioModel"
}

Full entity example with AlphaEdge

{
  "id": "audio-gen-model_xxxxxxxxx",
  "name": "AlphaEdge STT",
  "description": "AlphaEdge speech-to-text",
  "provider": "alphaedge",
  "config": {
    "connection": {
      "base_url": "https://api-endpoints.alphaedge-ai.com",
      "token": "${vault://local/ALPHAEDGE_API_KEY}",
      "timeout": 180000
    },
    "options": {
      "stt": {
        "enabled": true,
        "model": "alpha-audio-v1",
        "enable_diarization": true,
        "enable_postcorrect": true
      }
    }
  },
  "kind": "ai-gateway.extensions.cloud-apim.com/AudioModel"
}