Text to speech

POST /v1/audio/speech converts text to 24 kHz mono audio with Qwen 3 TTS (qwen3-tts).

Request

model (string, required) — qwen3-tts.
voice (string, required) — Voice id from List voices.
input (string, required) — Text to speak, up to 1000 characters.
response_format (string, optional, default wav) — wav, or pcm for headerless signed 16-bit little-endian samples at 24 kHz.
stream (boolean, optional, default false) — Writes the audio bytes as they are generated.

qwen3-tts rejects any other generation parameter, including speed, with 400 unsupported_parameter. The response body is audio/wav or audio/pcm. This example picks a featured voice that is ready for qwen3-tts and writes the streamed audio to speech.wav.

import os

import requests

headers = {"Authorization": f"Bearer {os.environ['HOPPER_API_KEY']}"}
voices = requests.get(
    "https://api.withhopper.com/voices?scope=featured", headers=headers
).json()["voices"]
voice_id = next(
    v["id"]
    for v in voices
    if any(p["model_id"] == "qwen3-tts" and p["status"] == "ready" for p in v["profiles"])
)

response = requests.post(
    "https://api.withhopper.com/v1/audio/speech",
    headers=headers,
    json={
        "model": "qwen3-tts",
        "voice": voice_id,
        "input": "Your appointment is confirmed for Thursday at ten.",
        "response_format": "wav",
        "stream": True,
    },
    stream=True,
)
with open("speech.wav", "wb") as f:
    for chunk in response.iter_content(chunk_size=4096):
        f.write(chunk)

import { writeFile } from "node:fs/promises";

const auth = { Authorization: `Bearer ${process.env.HOPPER_API_KEY}` };
const { voices } = await fetch(
  "https://api.withhopper.com/voices?scope=featured",
  { headers: auth },
).then((r) => r.json());
const voice = voices.find((v) =>
  v.profiles.some((p) => p.model_id === "qwen3-tts" && p.status === "ready"),
);

const response = await fetch("https://api.withhopper.com/v1/audio/speech", {
  method: "POST",
  headers: { ...auth, "Content-Type": "application/json" },
  body: JSON.stringify({
    model: "qwen3-tts",
    voice: voice.id,
    input: "Your appointment is confirmed for Thursday at ten.",
    response_format: "wav",
    stream: true,
  }),
});
await writeFile("speech.wav", Buffer.from(await response.arrayBuffer()));

VOICE_ID=$(curl -s "https://api.withhopper.com/voices?scope=featured" \
  -H "Authorization: Bearer $HOPPER_API_KEY" \
  | jq -r 'first(.voices[] | select(any(.profiles[]; .model_id == "qwen3-tts" and .status == "ready")) | .id)')

curl https://api.withhopper.com/v1/audio/speech \
  -H "Authorization: Bearer $HOPPER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "qwen3-tts",
    "voice": "'"$VOICE_ID"'",
    "input": "Your appointment is confirmed for Thursday at ten.",
    "response_format": "wav",
    "stream": true
  }' \
  --output speech.wav

List voices

GET /voices returns the voices your key can use, newest first.

scope (string, optional, default all) — featured for the voices Hopper publishes, mine for the voices your organization created.
limit (integer, optional, default 50) — Between 1 and 200.
cursor (string, optional) — next_cursor from the previous page.

curl "https://api.withhopper.com/voices?scope=featured&limit=10" \
  -H "Authorization: Bearer $HOPPER_API_KEY"

The response is {"voices": [...], "next_cursor": ...}, where each voice carries:

id (string) — Pass this as voice on /v1/audio/speech.
name (string) — Display name.
language (string) — Language of the voice.
profiles (array) — One {model_id, status} per model the voice is prepared for; the voice works with that model once status is ready.

GET /voices/{id} returns one voice. DELETE /voices/{id} deletes a voice your organization created.

Clone a voice

POST /voices/clone creates a voice from a single clip and returns 201 with the voice object. The request is multipart/form-data.

name (string, required) — Name for the new voice.
clip (file, required) — One file, wav, mp3, or flac, 5 to 30 seconds of one speaker. Clips with under 3 seconds of speech return 400 reference_audio_too_short.
transcript (string, optional) — What is said in the clip.
model_id (string, optional, default qwen3-tts) — Model to prepare the voice for.

curl https://api.withhopper.com/voices/clone \
  -H "Authorization: Bearer $HOPPER_API_KEY" \
  -F name="Support agent" \
  -F clip=@sample.wav \
  -F transcript="Thanks for calling, how can I help?" \
  -F model_id=qwen3-tts

Pass the returned id as voice on /v1/audio/speech. Error codes are listed in Errors.

​Request

​List voices

​Clone a voice

Request

List voices

Clone a voice