社区开发者 @xiaofenggan 用面壁智能 VoxCPM2 为 AI 短片生成角色配音，解决跨场景声音一致性问题。该工作流结合 Gemini 优化脚本与提示词、ChatGPT Image 2 生成视觉素材、Happyhorse 生成视频，实现角色音色与情绪风格稳定，减少配音返工。VoxCPM2 已开源，可帮助创作者构建更具辨识度的 AI 角色声音。
🎙️Build with VoxCPM2: Creating Consistent AI Character Voices for Short Films 🎬
A developer @xiaofenggan in our community created an AI short film with VoxCPM2 powering the character voice generation.🥳
In AI-generated content, keeping the same character voice across different scenes is still a challenge. This project explores how reference audio-based voice generation can help AI characters maintain a stable identity and more natural expression throughout a story.
With VoxCPM2, the project achieves:
🎙️ Voice consistency — keeping the same character voice across scenes
🎭 Stable expression — maintaining tone and emotional style
⚡ Efficient creation — reducing dubbing iterations for AI short films
The workflow combines:
🔹 Gemini for script and voice prompt optimization
🔹 ChatGPT Image 2 for visual assets
🔹 Happyhorse for video generation
🔹 VoxCPM2 for reference audio voice generation
🔹 Editing tools for final output
Beyond generating speech, VoxCPM2 enables creators to build AI characters with more recognizable voices and richer personalities.
A great example of how open-source speech models can empower the next generation of AIGC creators and storytelling experiences. 🚀
Watch the short film：
🔗 https://weixin.qq.com/sph/AhYK6wTnWr
来源：OpenBMB · x.com