MiniMax H3-Image Rerun at 16MP: Gap with MOSS Narrows on Same 15 Prompts
Aggressive-Park-373 · reddit · 2026-11-13
After being called out for running H3 at 4MP while MOSS used its full 4MP ceiling, the author reran the same 15 prompts with H3 generating at 16MP (up to 6272x2688) before downscaling to match display size. Faces and skin improved the most—pores, freckles and stray hairs survive; fine textures like foam and fabric also gain micro-detail. Composition, prompt adherence and text rendering are roughly unchanged, and UI mockups/charts still fail. Verdict: H3 was underrated at 4MP and now edges ahead on skin and material detail, though MOSS still reads better to most commenters.
More from Multimodal
- A 'very convincing' grape arbor photo shows how real AI images now look — misovalko · 2026-11-13
- Krea demos 'Agentic Editing' to SF growth crowd, calling it Cursor moment for creative work — angrypenguinPNG · 2026-11-13
- Manga ink backgrounds LoRA for NoobAI V-Pred delivers solid blacks and real linework — Historical_Song1864 · 2026-11-13
- ByteDance launches Seedance 2.5 Flash: 2x faster, ~1/3 the price, up to 30s video — lyraxana · 2026-11-13
- Claude Opus nails game 3D environment assets, but characters still go to Tripo and Meshy — chongdashu · 2026-11-13
- Mini-drama test: Seedance 2.5 outperforms Kling 4.0 in tension scenes, workflow shared — alecubudulecu · 2026-11-13