Ostris fine-tuned H3 Image VAE beats official version on nearly every metric
ostrisai · x · 2026-11-14
Developer ostrisai reports his fine-tuned H3 Image VAE decoder now beats the official P-H3-Image VAE on every metric except LPIPS, plus added Face ID and Text OCR metrics where it also leads (vs. MiniMax-H3-Image too). Per the shared chart, mean quality rank: Ostris 1.20, P-H3-Image 1.80, MiniMax 3.00 at the same 32:1 compression. Loss is still dropping, so training continues.
Related event: Indie Dev's Fine-Tuned H3 Image VAE Decoder Beats Official Version(2 posts)→
More from Multimodal
- Stripe residency dev draws all friends in JavaScript with GPT 6.2 Astra, results stun — gabrielchua · 2026-11-22
- Fable 5.2 plus WebGL demoed as a tool to "create anything" — hewarsaber · 2026-11-22
- Cinematic UFO folklore music video built with Suno and SLOPCANNON — DarkroomSignal · 2026-11-22
- Photographer crafts AI-generated digital muse Nous Girl on faux magazine cover — alexcovo_eth · 2026-11-22
- Turn any Google Maps location into a cinematic drone shot with AI — aziz4ai · 2026-11-22
- Open-source world model PAN 2 wows with GTA-like physics in driving demos — LinusEkenstam · 2026-11-22