Voice library

A voice is referenced in TTS requests as {"mode": "id", "id": "<voice_id>"}. Your organization sees two kinds of voices: platform voices from the shared library, which are read-only, and its own voices, created by cloning or localization. Editing or deleting a platform voice returns 403 voice_read_only; a voice outside your visibility returns 404 voice_not_found.

GET /voices

List and filter voices.

Param	Values	Default
`scope`	`featured`, `all`, `mine`, `saved`	`all`
`gender`	filter on the voice’s gender field	none
`language`	filter on language	none
`tag`	filter on tags	none
`q`	text search	none
`limit`	1–200	50
`cursor`	opaque cursor from a previous page	none

import requests

resp = requests.get(
    "https://api.withhopper.com/voices",
    headers={"Authorization": "Bearer sk_hopper_..."},
    params={"scope": "featured", "language": "en", "limit": 50},
)
voices = resp.json()["voices"]

const resp = await fetch(
  "https://api.withhopper.com/voices?scope=featured&language=en&limit=50",
  { headers: { Authorization: "Bearer sk_hopper_..." } },
);
const { voices, next_cursor } = await resp.json();

curl "https://api.withhopper.com/voices?scope=featured&language=en&limit=50" \
  -H "Authorization: Bearer sk_hopper_..."

The response is {"voices": [Voice, ...], "next_cursor": "..." | null}. Pass next_cursor back as cursor until it is null.

The voice object

Field	Type	Meaning
`id`	string	Voice id — use it in the TTS `voice` object
`name`	string	Display name
`description`	string	Free text
`language`	string	Voice language
`gender`	string	Voice gender
`tags`	string[]	Filterable labels
`featured`	boolean	In the curated platform set (`scope=featured`)
`source`	string	`library`, `instant_clone`, `pro_clone`, or `localized`
`parent_voice_id`	string	Set on localized voices — the voice it was derived from
`profiles`	Profile[]	Per-model preparations — see below
`is_owner`	boolean	Your organization owns this voice; mutation allowed
`saved`	boolean	You saved this voice (`scope=saved`)
`created_at`	string	Creation timestamp

Each profile records that the voice is prepared for one model: {model_id, profile_kind, profile_version, status, cached_at, expires_at}. status is pending, ready, or failed. A TTS request only succeeds against a model with a ready profile.

GET /voices/:id

Returns the Voice, or 404 voice_not_found.

PATCH /voices/:id

Body: {name?, description?, tags?} — name must be non-empty, tags a string array. Own voices only; returns the updated Voice.

DELETE /voices/:id

Soft-deletes an owned voice. Returns {"id": "...", "deleted": true}.

Save and unsave

POST /voices/:id/save returns {"id": "...", "saved": true}; DELETE /voices/:id/save returns saved: false. Saved voices appear under scope=saved. Any visible voice can be saved, including platform voices.

Voice profiles

POST /voices/:id/profiles prepares a voice for a model. OmniVoice uses in-context-learning (ICL) profiles built from the voice’s source audio. Body: {model_id?} — default omnivoice.

A cached profile already exists: 200 {"voice_id": "...", "status": "ready", "profile": {...}}. The voice is usable immediately.
No cached profile: 202 {"job_id": "...", "voice_id": "...", "status": "..."}. Poll GET /jobs/:id until the job succeeds.
The voice has no canonical source audio: 409 voice_source_missing.

A TTS request that names a voice without a ready profile for the target model returns 409 voice_profile_missing.

Get started

LLM

Text to Speech

Speech to Text

Platform

GET /voices

The voice object

GET /voices/:id

PATCH /voices/:id

DELETE /voices/:id

Save and unsave

Voice profiles

​GET /voices

​The voice object

​GET /voices/:id

​PATCH /voices/:id

​DELETE /voices/:id

​Save and unsave

​Voice profiles

GET /voices

The voice object

GET /voices/:id

PATCH /voices/:id

DELETE /voices/:id

Save and unsave

Voice profiles