Testing H3 T2V: struggles with eye color fidelity at medium distance
SIR_NVAX_A_LOT · reddit · 2026-08-22
User shares experience with the H3 text-to-video model, noting it struggles to capture accurate eye colors at a medium distance. The author is working on improving prompting for better adherence and展示了 results for 20 women and pairs of eyes.
More from Models
- Thinking Machines Launches Inkling Multimodal MoE Models on OpenRouter — soumithchintala · 2026-08-22
- Analysis: V4-Flash uses multi-cropping mechanism for fused visual representation — teortaxesTex · 2026-08-22
- TinyCast: Probabilistic Zero-Shot Forecasting for Embedded Devices — raws-labs · 2026-08-22
- User prefers Qwen3.6-35B-A3 over Qwen-3.8-27B on 3090: dumb but fast is better — DanGrover · 2026-08-22
- Nvidia paired Claude Opus 5 with memory and a supervisor to score 100% on ARC-AGI-3 — HaktanSuren · 2026-08-22
- The Batch: Grok 4.6 Arrives, Claude Watermarks, Qwen3.8 Max Open Weights — DeepLearningAI · 2026-08-22