面壁智能发布 baby-mynah 项目,基于 VoxCPM-0.5B 小模型实现可控语音生成。该项目通过 LoRA 专家模块支持 9 种英语口音的无缝融合与独特音色创造,展示低成本微调下的高效语音定制能力,并已开放 demo 与模型权重。
🐦⬛ Meet baby-mynah — a tiny TTS bird powered by VoxCPM-0.5B.
Inspired by mynah birds that mimic different voices, this project explores controllable voice generation with small TTS models.
Built on VoxCPM-0.5B, the project uses LoRA experts for 9 English accents to create smooth accent morphing and unique voices.
This project demonstrates the potential of small TTS models in:
🎙️ Efficient voice customization with low-cost fine-tuning
🌍 Accent control and voice style exploration
🪶 Creative on-device and lightweight TTS applications
A great example of how compact open-source models can enable more accessible and creative AI experiences.
Try the demo:
🔗 https://huggingface.co/spaces/build-small-hackathon/baby-mynah
来源:OpenBMB · x.com