GET
/v1/schemas/byteplus/audio/tts/create
Accept:
| field | recorded source |
|---|---|
| schema | Source linked above. probe-verified Source traceebffeaf934197130d6b4cc6534a33b6b04d51a5a71ce2403b2bcf8c7e2cb627f{
"fetchedAt": 1786770726,
"extractorVersion": "1",
"verifiedAt": "2026-08-12"
} |
Request body
5 properties · 2 required · 2 $defs| property | type | description | constraints |
|---|---|---|---|
| model* | string | Seed Speech synthesis model, e.g. seed-audio-1.0. | |
| text_prompt* | string | The text to speak. Dual-purpose: literal text or a natural-language description of the delivery, and where the @Audio1..@Audio3 reference markers go. (The field is text_prompt, not text — the 'text' spelling belongs to the separate /tts/unidirectional endpoint… | length ≤ 3000 |
| references | TTSReference[] | ||
| audio_config | TTSAudioConfig | Audio output configuration. The *_rate fields are integer percentages relative to the voice's neutral delivery, not multipliers. | |
| watermark | object | Watermark the generated audio. NOT a boolean: `true` and a bare string both fail with 500 `55000000 kiteX processing … thrift marshal`, while an object is accepted with 200 — so the wire type is a struct. Its member names could not be confirmed (an unrecognise… |
Types
2 $defsTTSAudioConfigobject
Audio output configuration. The *_rate fields are integer percentages relative to the voice's neutral delivery, not multipliers.
| format | enum | Output container/codec. Server default 'wav'. | wav | mp3 | pcm | ogg_opus |
| sample_rate | integer | Output sample rate in Hz. The docs list 8000/16000/24000/32000/44100/48000, but the endpoint is not strict — 40000 (the documented default, absent from that list) is accepted with 200. Left unconstrained rather than enumerated so validation does not reject req… | |
| speech_rate | integer | Speaking rate: -50 = 0.5x, 0 = 1x, 100 = 2x. | -50 ≤ n ≤ 100 |
| loudness_rate | integer | Loudness adjustment; 0 is the voice's natural level. | -50 ≤ n ≤ 100 |
| pitch_rate | integer | Pitch adjustment; 0 is the voice's natural pitch. | -12 ≤ n ≤ 12 |
| enable_subtitle | boolean | Emit sentence and word timings in the response. Default false. |
TTSReferenceobject
Voice selection / cloning reference. FLAT shape — the members sit directly on the entry, with no discriminating `type` field (an entry carrying one is accepted only because unknown keys are ignored); an unknown speaker id fails 400 45001115 'speaker … not foun…
| speaker | string | Stock voice id, e.g. en_female_stokie_uranus_bigtts. | |
| audio_url | string | URL of a reference clip to clone (≤30s, ≤10MB). | |
| audio_data | string | Base64 reference clip to clone (≤30s, ≤10MB). | |
| image_url | string | URL of a reference image (≤10MB). | |
| image_data | string | Base64 reference image (≤10MB). |
Validate a payload
before you spend tokensshell
$ curl -X POST https://modelschemas.com/v1/validate \
-d '{"provider":"byteplus","endpointId":"tts/create","payload":{…}}'
→ {"valid": false, "errors": [{"path": "#", "keyword": "required", …}]}