Shizuku AI unveils in-house singing lip sync tech, beating uLipSync in quality
bdsqlsz · x · 2026-11-01
Shizuku AI's tech team shared its first post introducing the proprietary lip sync technology used by its virtual character Shizuku when singing.
- Developed in-house, the approach reportedly achieves noticeably better lip sync quality than the commonly used uLipSync, with smoother results in singing scenes.
- A comparison video shows Ground Truth vs. their 1D CNN method vs. uLipSync side by side.
- The team says it will keep sharing technical details of Shizuku's development going forward.
More from Fun
- macOS Local Network permission list shows dozens of duplicate Google Chrome entries — jwt0625 · 2026-11-01
- Meme: writing an LLVM backend for a fully undocumented GPU architecture to ditch closed-source nvcc — mike64_t · 2026-11-01
- Dimillian's Codex goes off the rails, pulls up Backrooms images as game references — Dimillian · 2026-11-01
- ML author Burkov mocks "AI will kill all jobs" with farming history irony — burkov · 2026-11-01
- Unreleased game engine hits 120 FPS on an iPhone, dev stunned — Dimillian · 2026-11-01
- WoW players meme Altman's 'your RAM belongs in data centers' take — max_paperclips · 2026-11-01