Yapper API Docs

POST /api/v1/processes with "type": "audio-generation". Pick any model — the API does not plan-gate models. GET /api/v1/models returns the same capabilities at runtime.

Input fields

Field	Type	Required	Description
`prompt`	`string`	yes	What to generate.
`voiceId`	`string`	no	A preset voice id returned by GET /api/v1/audio/voices?modelId=... for the selected model.
`referenceImages`	`({assetId} \| {url})[]`	no	Image references. Send either {assetId} for a Yapper asset or {url} for a public URL to import. Legacy {docId, url} objects remain accepted.
`referenceAudios`	`({assetId} \| {url})[]`	no	Audio references. Send either {assetId} for a Yapper asset or {url} for a public URL to import. Legacy {docId, url} objects remain accepted; private storage metadata is resolved server-side.
`speed`	`number \| string`	no	Speech speed multiplier when supported.
`language`	`string`	no	Requested spoken language code when supported.
`volume`	`number`	no	Output volume multiplier when supported.
`emotionString`	`string`	no	Provider-specific emotion or delivery direction.
`pitch`	`number`	no	Pitch adjustment in semitones when supported.
`multilingual`	`boolean`	no	Enable the model's multilingual mode for mixed-language prompts.
`disableAutoRetries`	`boolean`	no	Disable automatic same-model retries. Defaults to false, so automatic retries are enabled.

Referencing media: referenceImageUrls takes plain https image URLs. The normalized referenceImages, referenceVideos, and referenceAudios fields accept either {assetId} for an existing Yapper asset or {url} for public media, which is imported automatically. Provider-native asset:// URIs are internal and are not accepted here. Legacy {docId, url} objects remain accepted for backwards compatibility.

Example

Add "dryRun": true to get the exact credit cost without starting, and always send an Idempotency-Key header on real starts.

POST /api/v1/processes
{
  "type": "audio-generation",
  "model": "sonic-3.5",
  "input": {
    "prompt": "Read this in a warm, conversational tone.",
    "voiceId": "voice_id"
  }
}

Models at a glance

Credits are per generated audio output (exact prices in each model's section).

Model	Company	Preset voices	Text only	Ref audio	Image ref
`sonic-3.5`	cartesia	yes	no	1	no
`eleven_v3`	elevenlabs	yes	no	0	no
`bytedance/seed-audio-1.0`	bytedance	yes	yes	3	yes

Sonic 3.5 (sonic-3.5)

Cartesia's latest natural, low-latency speech model with support for 42 languages.


Credits	0
Reference images	up to 0
Reference audio	up to 1
Preset voices	supported
Text-only generation	not supported
Image reference	not supported
Speed	0.6–1.5
Volume	0.5–2
Audio output	yes

Eleven v3 (eleven_v3)

ElevenLabs' most expressive speech model with audio-tag control and support for 70+ languages.


Credits	0
Reference images	up to 0
Reference audio	up to 0
Preset voices	supported
Text-only generation	not supported
Image reference	not supported
Audio output	yes
Max prompt length	5000

Seed Audio 1.0 (bytedance/seed-audio-1.0)

Generate speech, sound effects, music, and atmosphere from text and multimodal references.


Credits	0
Reference images	up to 1
Reference audio	up to 3 (90s total)
Preset voices	supported
Text-only generation	supported
Image reference	supported
Speed	0.5–2
Volume	0.5–2
Pitch	-12–12
Audio output	yes
Max prompt length	2048

Audio generation