# Speech
> AI text-to-speech — turn text into natural studio-grade audio, pay per clip in USDC or EURC. No API key, no account required.

Text-to-speech behind x402 micropayments. A single POST /speech/synthesize endpoint accepts up to 4,000 characters of text plus a voice and returns synthesized audio. Powered by the qwen-audio-3.0-tts-plus model (premium speech tier of the Qwen Audio family) with two flagship voices: longanlufeng (male, bright and cheerful) and longanlingxin (female, warm and empathetic), each fluent in Chinese (Mandarin) and English. Output formats: mp3 (default), wav, pcm, opus, with volume, speech-rate and pitch controls. Clips are stored for 2 hours after generation — poll GET /speech/status/{audio_id} (free) and download via GET /speech/audio/{audio_id} (free).

## Pricing
- `exclusive` tier — **$0.256125** on Base: `/speech/synthesize`
- Chains: Base, Arbitrum, Polygon, Avalanche. Per-chain prices in `/.well-known/x402.json`.

## Operations
### inference
- **POST /speech/synthesize** — Synthesize speech from text on the qwen-audio-3.0-tts-plus model. Billed on the Exclusive tier, per clip. Returns an audio_id immediately; poll /speech/status/{audio_id} then download /speech/audio/{audio_id} (both free). — params: `text` (string)*, `voice` (string), `format` (string), `sample_rate` (integer), `volume` (integer), `rate` (number), `pitch` (number) (tier: `exclusive`)

## Workflows
- **speech_synthesize** (Turn text into audio: POST /speech/synthesize (paid) → poll GET /speech/status/{audio_id} (free) until status is completed → GET /speech/audio/{audio_id} (free) for the clip): `synthesize`

## Data sources
- Alibaba DashScope API (dashscope-intl.aliyuncs.com) — qwen-audio-3.0-tts-plus via the native SpeechSynthesizer endpoint. Proxied server-side; the service's own API key is never exposed to callers. Upstream audio URLs are short-lived — the service downloads and re-serves the bytes.

## Free endpoints
- GET `/`, `/health`, `/about`, `/.well-known/x402.json`, `/llms.txt`, `/openapi.json` (standard)
- GET `/speech/voices` — List available voices with characteristics and accepted formats/ranges
- GET `/speech/status/{audio_id}` — Poll a synthesis: synthesizing | completed | failed (with error) | not_found
- GET `/speech/audio/{audio_id}` — Download a completed clip — free, retained 2 hours
- GET `/speech/feedback` — Return the 50 most recent feedback entries
- POST `/speech/feedback` — Submit feedback (response_id, rating 1-5, optional comment)

## Examples
- `synthesize` — Speak a short English announcement with the male flagship voice: `{"text": "Welcome to the x402 speech service.", "voice": "longanlufeng", "format": "mp3"}`