# Timbre by Uploading > Timbre is a voice cloning and text-to-speech API by Uploading: streaming speech in ten languages, zero-shot cloning from a few seconds of audio, OpenAI-SDK compatible, priced at $10 per million UTF-8 bytes of input text. ## Product facts - Languages: Chinese (Mandarin), English, Japanese, Korean, German, French, Spanish, Italian, Portuguese, Russian. - Price: $10 per 1,000,000 UTF-8 bytes of text that produces audio; failed requests are not charged. Packs: $10 = 1M bytes, $50 = 5M bytes, $200 = 21M bytes (+5%). 100,000 free bytes on sign-up. - Bytes to speech (measured on Timbre's voices): 1M bytes ≈ 22 hours of Chinese or 18 hours of English speech. An English letter is 1 byte, a Chinese character 3 bytes. - Speed: median first audio about 59 ms on the demo library takes; 75 ms p95 with 16 concurrent streams per GPU in Uploading's load test. - Model today: Qwen3-TTS 1.7B (Apache-2.0), served by Uploading in the US (Iowa) and Tokyo. - Data: customer text, audio and voices are never used for training; voices can be deleted; cloning requires the speaker's consent. - Status: the API is opening by request; the live demo is free. Terms, privacy and acceptable use are drafts, not yet in force. ## Links - [Product and pricing](https://www.uploading.tech/voice#pricing) - [API docs](https://www.uploading.tech/voice/docs): POST /v1/tts, OpenAI-compatible POST /v1/audio/speech (model "timbre-1"), voices, wallet, errors, limits - [Console / request access](https://www.uploading.tech/voice/console) - [Terms (draft)](https://www.uploading.tech/voice/terms), [Privacy (draft)](https://www.uploading.tech/voice/privacy), [Acceptable use (draft)](https://www.uploading.tech/voice/acceptable-use) - Contact: info@uploading.tech · abuse reports: abuse@uploading.tech