---
title: VoiceCraft CPU Clone Server
emoji: 🎙️
colorFrom: gray
colorTo: blue
sdk: docker
app_port: 7860
---

# VoiceCraft CPU Clone Server

Use this README content for CPU/Docker Spaces.

Settings:

- `VOICECRAFT_AUTH_MODE=license`
- `LICENSE_VALIDATION_URL` - deployed Google Apps Script URL
- `ENABLE_CLONE_ENGINES=1`
- `HF_TOKEN` - Hugging Face read token from an account that accepted the Pocket TTS model terms.

`API_SECRET` is legacy compatibility only and should not be distributed in desktop clients. Use `VOICECRAFT_AUTH_MODE=layered` only during migration, then rotate/remove the old shared secret.

Voice clone access:

1. Open `https://huggingface.co/kyutai/pocket-tts`.
2. Accept the model terms/access form.
3. Create a read token at `https://huggingface.co/settings/tokens`.
4. Add it to this Space as `HF_TOKEN`.
5. Restart or rebuild the Space.

In the Google Sheet row for this Space:

- `AppLink` = this Docker Space URL
- `VoiceClone` = `active`, `enabled`, `on`, or `true`
- Configure `DeviceID` and `MaxDevices` for device binding.

This Space serves `/tts`, `/health`, `/status`, and `/all_voices`. `/tts` validates the license and registered device server-side; clone access also requires the row's `VoiceClone` entitlement. Voice cloning runs on CPU through Pocket TTS and does not use ZeroGPU quota.

Voice Clone accepts long scripts from the desktop app. For best speed, keep each generation under 10,000 characters; longer scripts are batched automatically and joined into one WAV.