Voices

Every synthesis request takes a voice_id. Chariot provides a catalog of ready-to-use voices across English and Indian languages and accents.

Listing voices

GET /v1/voices returns the voices available to your workspace.

curl "https://api.chariot.in/v1/voices?limit=100" \
  -H "chariotai-api-key: YOUR_API_KEY"

Each voice includes its identity, preview URLs, per-model capabilities, and availability. This abbreviated example illustrates the response shape; fetch the catalog for current values:

{
  "total_voices": 8,
  "data": [
    {
      "voice_id": "bac7d666-094d-4698-91fa-741d60fce662",
      "voice_name": "Matt",
      "accent": "american",
      "gender": "male",
      "models": [
        {
          "model": "chariot-multilingual-v1",
          "label": "Chariot Multilingual v1",
          "samples": {
            "en": "https://example.com/preview-en.wav"
          },
          "speeds": [1.0],
          "max_chars": 1000
        }
      ],
      "status": "published",
      "status_reason": null,
      "restricted": false,
      "restriction_reason": null,
      "restricted_on_change": false
    }
  ]
}

Models and previews

Choose an entry in models and pass its model value as model_type in your TTS request. Use that entry’s speeds to select a speaking rate. Preview URLs and speaking rates are model-specific. Read models[].samples and models[].speeds; the response does not include top-level voice_sample or available_speeds fields. models[].samples maps the language of the preview text to its audio URL. For example, en identifies a sample generated from English text. It does not identify the speaker’s actual language or limit which languages you can synthesize. A model can have no preview samples. models[].max_chars is the limit for HTTP streaming and each WebSocket fragment and unflushed buffer. Sync has separate model limits. Use a voice only when status is published and restricted is false. Other voices return an empty models list. status_reason explains a failure; restriction_reason explains a current restriction. restricted_on_change describes a scheduled restriction and does not affect use today.

Filtering

Narrow the catalog with query parameters. Combine them freely:

Parameter	Values
`accent`	`american`, `british`, `indian`
`gender`	`male`, `female`, `neutral`
`voice_name`	Case-insensitive substring match
`q`	Name-search alias; takes precedence over `voice_name` when nonempty
`status`	Comma-separated `draft`, `processing`, `failed`, `published`; deleted voices are never returned
`limit`	1 to 500, default 10
`skip` / `offset`	Pagination offset, default 0; a nonzero `offset` takes precedence
`sort`	`desc` (default), `asc`

curl "https://api.chariot.in/v1/voices?accent=indian&gender=female" \
  -H "chariotai-api-key: YOUR_API_KEY"

The response includes total_voices and an X-Total-Count header with the matching count before pagination.

Available voices

Name	Accent	Gender
Matt	American	Male
Mia	American	Female
Ryan	British	Male
Karan	Indian	Male
Darshan	Indian	Male
Meera	Indian	Female
Neha	Indian	Female

Voice ids are stable, but the catalog grows over time. Fetch GET /v1/voices at runtime rather than hardcoding the full list.

Get started

Text to speech

Speech to text

Resources

Listing voices

Models and previews

Filtering

Available voices

​Listing voices

​Models and previews

​Filtering

​Available voices

Listing voices

Models and previews

Filtering

Available voices