Clone your voice in seconds, then have it read any text in real time.
Record your voice first to unlock this step.
No training, no waiting. Your recording never trains a model.
Type it. Hear it as it’s made.
Optional: heard once before you hear it.
Read one line. That’s your voice.

First sentence to last, one person.
One price. Streaming. OpenAI-compatible.
Concurrent requests scale with your total top-ups: 5 to start, 15 from $100, 50 from $1,000. Pricing
from openai import OpenAI
client = OpenAI(base_url="https://api.uploading.tech/v1", api_key=TIMBRE_API_KEY)
with client.audio.speech.with_streaming_response.create(
model="timbre-1", voice="VOICE_ID", input="Hello from Timbre.",
) as r:
r.stream_to_file("hello.mp3")
curl https://api.uploading.tech/v1/audio/speech \
-H "Authorization: Bearer $TIMBRE_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model": "timbre-1", "voice": "VOICE_ID", "input": "Hello from Timbre."}' \
--output hello.mp3
import OpenAI from "openai";
const client = new OpenAI({ baseURL: "https://api.uploading.tech/v1", apiKey: process.env.TIMBRE_API_KEY });
const r = await client.audio.speech.create({ model: "timbre-1", voice: "VOICE_ID", input: "Hello from Timbre." });
await (await import("node:fs/promises")).writeFile("hello.mp3", Buffer.from(await r.arrayBuffer()));
By the text that produces audio: $10 per million UTF-8 bytes of text. Failed requests are free.
Yes. Read one sentence on this page to clone your voice right in your browser.
No. Demo recordings stay in memory for 15 minutes, are never written to disk, and are never used for training.
Your own, or that of someone who has agreed to how you’ll use it. Impersonating anyone to deceive is not allowed.
Chinese, English, Japanese, Korean, German, French, Spanish, Italian, Portuguese and Russian.
Yes. Set the SDK's base_url to https://api.uploading.tech/v1 and use a Timbre key.