GET
/v1/schemas/fal/audio/fal-ai%2Fmaya%2Fstream
Accept:
Request body
8 properties · 2 required · 0 $defs| property | type | description | constraints |
|---|---|---|---|
| prompt* | string | Description of the voice/character. Includes attributes like age, accent, pitch, timbre, pacing, tone, and intensity. See examples for format. | length ≤ 500 |
| text* | string | The text to synthesize into speech. You can embed emotion tags anywhere in the text using the format <emotion_name>. Available emotions: laugh, laugh_harder, sigh, chuckle, gasp, angry, excited, whisper, cry, scream, sing, snort, exhale, gulp, giggle, sarcasti… | length ≤ 5000 |
| temperature | number | Sampling temperature. Lower values (0.2-0.5) produce more stable/consistent audio. Higher values add variation. | 0 ≤ n ≤ 2 · default: 0.4 |
| sample_rate | enum | Output audio sample rate. 48 kHz uses upsampling for higher quality audio, 24 kHz is native SNAC output (faster, lower latency). | 48 kHz | 24 kHz · default: "24 kHz" |
| repetition_penalty | number | Penalty for repeating tokens. Higher values reduce repetition artifacts. | 1 ≤ n ≤ 2 · default: 1.1 |
| output_format | enum | Output audio format. 'mp3' for browser-playable audio, 'wav' for uncompressed audio, 'pcm' for raw PCM (lowest latency, requires client-side decoding). | mp3 | wav | pcm · default: "mp3" |
| max_tokens | integer | Maximum number of SNAC tokens to generate (7 tokens per frame). Controls maximum audio length. | 28 ≤ n ≤ 4000 · default: 2000 |
| top_p | number | Nucleus sampling parameter. Controls diversity of token selection. | 0 ≤ n ≤ 1 · default: 0.9 |
Validate a payload
before you spend tokensshell
$ curl -X POST https://modelschemas.com/v1/validate \
-d '{"provider":"fal","endpointId":"fal-ai/maya/stream","payload":{…}}'
→ {"valid": false, "errors": [{"path": "#", "keyword": "required", …}]}