/venice-audio-speech

name: venice-audio-speech
description: Generate speech from text via POST /audio/speech, and clone a voice via POST /audio/voices. Covers TTS models (Kokoro, Qwen 3, xAI, Inworld, Chatterbox, Orpheus, ElevenLabs Turbo, MiniMax, Gemini Flash, Gradium), voices per model, cloned-voice handles and raw ElevenLabs Voice IDs, per-model output formats (model_spec.supported_formats / default_format), streaming, prompt/style control, temperature/top_p, speed clamping, and language hints.

Venice TTS (`/audio/speech`)

`POST /api/v1/audio/speech` converts text to an audio stream or file. OpenAI-compatible — the OpenAI SDK's `audio.speech.create()` works as a drop-in.

| Method | Path | Auth | Notes | |---|---|---|---| | `POST` | `/api/v1/audio/speech` | Bearer key or x402 (SIWX) | JSON body, returns raw audio. Billed per input character. | | `POST` | `/api/v1/audio/voices` | Bearer key or x402 (SIWX) | `multipart/form-data`. Clone a voice → `vv_…` handle. There is no `GET /audio/voices`. |

Use when

• You want narration, voice replies, or UI audio from text.
• You need a specific voice family (ElevenLabs, Kokoro, xAI, Qwen 3, Orpheus, Chatterbox, MiniMax, Inworld, Gemini Flash, Gradium).
• You want streaming audio returned as it is generated.
• You need style/emotion control on supported models, or synthesis in a cloned voice.

For music, sound effects and the character-priced ElevenLabs TTS models — v3, v4, v4 Turbo, Multilingual v2 — (async), see [`venice-audio-music`](../venice-audio-music/SKILL.md). For transcription (audio → text), see [`venice-audio-transcription`](../venice-audio-transcription/SKILL.md).

Minimal request

curl https://api.venice.ai/api/v1/audio/speech \
  -H "Authorization: Bearer $VENICE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "tts-xai-v1",
    "voice": "eve",
    "input": "Hello, welcome to Venice Voice.",
    "response_format": "mp3",
    "speed": 1.0,
    "streaming": false
  }' --output hello.mp3

Response is the raw audio. `Content-Type` is `audio/mpeg`, `audio/opus`, `audio/aac`, `audio/flac`, `audio/wav` or `audio/pcm` depending on the format actually produced.

Request schema

The body is strict — unknown fields return `400`.

| Field | Type | Default | Notes | |---|---|---|---| | `input` | string | — | **Required.** 1–4096 characters, must contain non-whitespace. Text that sanitizes to nothing speakable (e.g. only markup/emoji) → `400 "Input must contain speakable text"`. | | `model` | string | — | **Send it.** The OpenAPI schema lists a `tts-kokoro` default, but that default is never applied: omitting `model` returns `400 "model is required"`. Unknown id → `404`. | | `voice` | string, ≤ 512 | the model's default voice | Voices are model-specific; a voice from another model → `400`. Also accepts a cloned-voice handle (`vv_…`) from `POST /audio/voices` (same `model` that created it), and — on models with `supports_custom_voice_id: true` (currently `tts-elevenlabs-turbo-v2-5`) — a raw provider Voice ID. | | `response_format` | `mp3` / `opus` / `aac` / `flac` / `wav` / `pcm` | the model's `default_format` | **Support is per model** — read `model_spec.supported_formats` / `default_format`. Requesting a format the model doesn't support → `400`. | | `speed` | number | `1.0` | Schema range `0.25–4.0`. Passed to Kokoro unchanged; clamped by xAI (`0.7–1.5`), ElevenLabs Turbo (`0.7–1.2`) and MiniMax (`0.5–2`); ignored by the other models. | | `streaming` | bool | `false` | `true` → chunked audio stream as it's generated. `false` → buffered file with `Content-Length`. | | `language` | string, 2–32 chars | — | Optional hint; form is model-specific (see below). Gemini Flash drops values outside its locale list and ElevenLabs Turbo drops values longer than 5 chars; Qwen 3, MiniMax and xAI forward the value as given, and a value the model rejects returns `400 "Invalid request parameters: language"`. Other models ignore it. | | `prompt` | string, ≤ 500 | — | Style/emotion instruction. Used by Qwen 3 and Gemini Flash (sent as style instructions); ignored elsewhere. | | `temperature` | number, 0–2 | — | Used by Qwen 3, Orpheus, Chatterbox HD, Gemini Flash; ignored elsewhere. | | `top_p` | number, 0–1 | — | Qwen 3 only; ignored elsewhere. |

`language` by model: Qwen 3 → full names (`English`, `Chinese`, …; default auto); xAI → ISO 639-1 (`en`), passed through as given, so send a valid code (default auto); ElevenLabs Turbo → ISO 639-1 (values longer than 5 chars dropped, shorter ones forwarded); MiniMax → full names (sent as a language boost, unchecked); Gemini Flash → full locale strings such as `English (US)` or `Japanese (Japan)` (anything else dropped). Kokoro, Inworld, Chatterbox, Orpheus and Gradium ignore it.

Models

Every id below is in the live `GET /models?type=tts` list. Prices are `model_spec.pricing.input.usd` = **USD per 1M input characters**.

| Model ID | Default voice | Formats (default first) | Privacy | Notes | |---|---|---|---|---| | `tts-kokoro` | `af_sky` | mp3, opus, aac, flac, wav, pcm | private | Multilingual via voice prefix. `speed` passed through unclamped. | | `tts-qwen3-0-6b` / `tts-qwen3-1-7b` | `Vivian` | mp3 | private | `prompt`, `temperature`, `top_p`, `language`. | | `tts-xai-v1` | `eve` | mp3, wav, pcm | anonymized | 26 voices, ISO `language`. | | `tts-inworld-1-5-max` | `Craig` | wav | anonymized | Low-latency; all voices are English. | | `tts-chatterbox-hd` | `Aurora` | wav | private | `temperature`. **Voice cloning (zero-shot).** | | `tts-orpheus` | `tara` | wav | private | `temperature`. | | `tts-elevenlabs-turbo-v2-5` | `Rachel` | mp3 | anonymized | Accepts raw ElevenLabs Voice IDs as `voice`. | | `tts-minimax-speech-02-hd` | `WiseWoman` | mp3, pcm, flac | anonymized | `language`. Cloning with this model is **not available** to regular keys (see below). | | `tts-gemini-3-1-flash` | `Kore` | mp3, opus, wav | anonymized | `prompt`, `temperature`, locale `language`. | | `tts-grad