Repository Issue Activity (beta)

#59 Audio.cpp now supports LiveAvatar. C++/GGML. No Python. Generate 720p video with 16GB VRAM.

#58 Block discontinuity after Stage 1 causal training

#57 实时回复时间过长

#55 实时交互的部分找不到

#52 AI Interview Coach

#6 训练代码发布

#53 二维码不可用

#54 群二维码超过200人，加不进去

#47 二维码又不可用了

#48 About Benchmark

#50 虚拟人形象自定义

#49 body gesture

#38 FP8 + ENABLE_COMPILE on L40s (48GB): Expected Support and Performance vs H100?

#45 Can the WeChat group join QR code be resent?

#42 release plan of training code.

#43 Recent Advancements Surpassing LiveAvatar and Future Plans

#41 Request for timeline on 4-GPU, 4-step inference version release

#39 Any plan to develop 3-step version?

#40 关于生成视频姿态幅度过大

#37 About training details

#36 ~4s First-Frame Delay Every Time TTS Audio Is Sent (Not Only Cold Start) with torch.compile + FP8

#35 Low FPS (~6) with block-wise streaming on 5×H100 (80GB) using FP8 + offload_model=False

#34 Clarification on `ENABLE_FP8` environment variable vs `--fp8` argument

#27 High Latency in Streaming Pipeline Due to VAE Decode–Reencode Loop | Seeking Guidance on Latent Reuse / Faster Reference Update

#33 Need help for block-wise decoding implementation using single GPU