Minimax H3 Ref2VA Issue: Ignoring Reference Images
Thorozar · reddit · 2026-08-30
The author reports an issue with Minimax H3's Ref2VA feature where the model sometimes ignores reference images entirely, treating them as if they don't exist. Copying exact subject and retention sections from working prompts did not resolve the issue. The author is seeking solutions from the community.
More from Apps
- Autonomous Lamp: a $999 open-source desk robot lamp you can customize — dee_hw · 2026-10-15
- Perplexity's Photon post reveals how to rank: a playbook for LLM visibility — gaganghotra_ · 2026-10-15
- ChatGPT Ate Search — Is the Calendar Next? — gabe_ragland · 2026-10-15
- Inanimate preps WORKS: AI-agent-powered physical devices you control with prompts — genmon · 2026-10-15
- Adaption launches Visualize, first release of its Adaptive Interfaces AI system — yuntiandeng · 2026-10-15
- ChatGPT ranks its own Dots first in personal agent comparison, but user reviews tell otherwise — vista8 · 2026-10-15