Skip to content

Transcription Models

Requires Authentication

Choose a transcription model for each transcription or streaming request. Its identifier remains stable when administrators update the underlying model version.

Sentence verification (sentver) models are documented separately in Sentence Verification Models. Sentver models are not returned by this transcription model API.

Model Types

A transcription model can be one of three types:

  • FAMILY: A high-level reference that always points to the latest recommended version of a transcription model family (e.g., large-v3).
  • PINNED: A reference that points to a specific, stable version of a transcription model.
  • CONCRETE: A direct reference to a specific transcription model version (primarily used for internal tracking).

GET /api/v1/speech-services/models

Requires Authentication - Scopes: speech:model-aliases:read

Returns a list of all transcription models accessible to the current user. These are the identifiers you use in transcription and streaming requests. isDefault identifies the default transcription model. Clients can use it to initialize model selection, then send the chosen model ID in the request's required model field.

ModelAliasListResponse

This endpoint returns an array of ModelAliasResponse objects.

ModelAliasResponse
Field Type Nullable Description
id string No Transcription model identifier (e.g., large-v3).
type string enum No Model type. One of: FAMILY, PINNED, CONCRETE.
displayName string No Human-readable name of the transcription model.
description string Yes Brief description of the transcription model's purpose.
status string enum No Availability status. One of: AVAILABLE, UNAVAILABLE, MAINTENANCE, DEPRECATED.
currentVersion string No Current resolved transcription model version name.
isDefault boolean No Whether this is the default transcription model.
capabilities ModelCapabilities No Technical capabilities and supported features.
deprecated boolean No Whether this model is deprecated.
deprecationDate string (ISO 8601) Yes Date when the model was marked as deprecated.
deprecationMessage string Yes Information regarding deprecation or migration instructions.
ModelCapabilities
Field Type Nullable Description
languages string[] No List of supported BCP-47 language codes.
supportsLangAutodetect boolean No Whether the transcription model supports automatic language detection.
supportsStreaming boolean No Whether the transcription model supports real-time streaming transcription.
supportsBatch boolean No Whether the transcription model supports batch (offline) transcription.
supportsPunctuation boolean No Whether the engine provides punctuation and capitalization.
supportsDiarization boolean No Whether the engine provides speaker diarization.
supportsItn boolean No Whether the engine provides inverse text normalization.
ModelAliasResponse
{
  "id": "large-v3",
  "type": "FAMILY",
  "displayName": "Large V3 (Multilingual)",
  "description": "Our most accurate multilingual model for general purpose transcription.",
  "status": "AVAILABLE",
  "currentVersion": "v3.0.2",
  "isDefault": true,
  "capabilities": {
    "languages": ["en", "he", "fr", "de", "es"],
    "supportsLangAutodetect": true,
    "supportsStreaming": true,
    "supportsBatch": true,
    "supportsPunctuation": false,
    "supportsDiarization": false,
    "supportsItn": false
  },
  "deprecated": false,
  "deprecationDate": null,
  "deprecationMessage": null
}
Status Description
200 OK List of available transcription models returned.
401 Unauthorized Authentication required or token invalid.
403 Forbidden Insufficient permissions (missing speech:model-aliases:read scope).

GET /api/v1/speech-services/models/{model}

Requires Authentication - Scopes: speech:model-aliases:read

Returns detailed information for a specific transcription model.

Path Parameters
Field Type Required Description
model string Yes The name of the transcription model.
ModelAliasResponse

Returns the same structure as described in the list endpoint.

Status Description
200 OK Transcription model details returned.
401 Unauthorized Authentication required or token invalid.
403 Forbidden Insufficient permissions.
404 Not Found Transcription model not found.

GET /api/v1/speech-services/models/{model}/languages

Requires Authentication - Scopes: speech:model-aliases:read

Returns a list of supported languages for the specified transcription model.

Path Parameters
Field Type Required Description
model string Yes The name of the transcription model.
ModelLanguagesResponse
Field Type Nullable Description
model string No The transcription model name.
supportsLangAutodetect boolean No Whether the transcription model supports automatic language detection.
languages string[] No List of supported BCP-47 language codes (e.g., en, he).
ModelLanguagesResponse
{
  "model": "large-v3",
  "supportsLangAutodetect": true,
  "languages": ["en", "he", "fr", "de", "es", "it", "ja", "ko"]
}
Status Description
200 OK Supported languages returned successfully.
401 Unauthorized Authentication required or token invalid.
403 Forbidden Insufficient permissions.
404 Not Found Transcription model not found.