README.md

metadata

title: VoxCPM Demo LX
emoji: 🎙️
colorFrom: blue
colorTo: indigo
sdk: gradio
sdk_version: 6.0.0
app_file: app.py
python_version: '3.10'
pinned: false
license: apache-2.0

Lightweight Hugging Face Spaces demo for VoxCPM. It currently defaults to a VoxCPM1.5 placeholder model until VoxCPM2 weights are available.

Notes:

• The first generation request may spend extra time downloading/loading the model.
• Ultimate Cloning Mode requires both a reference audio clip and a usable transcript.
• While the default placeholder is VoxCPM1.5, reference audio works through transcript-based prompting rather than VoxCPM2 reference-audio cloning.
• Control Instruction is effectively a VoxCPM2-only option and will be ignored by the placeholder model.

Environment variables:

• HF_REPO_ID: model repo id, defaults to openbmb/VoxCPM1.5
• VOXCPM_MODEL_DIR: optional local checkpoint directory override