Local models fall short of nano banana 2 for multi-reference face scene generation
classified18 · reddit · 2026-10-14
A Reddit user asks how to replicate nano banana 2's multi-reference face scene generation locally, after trying krea 2, flux 2 dev, flux 2 klein 9b, and qwen image edit 2.1 without coming close to the paid service's quality.
They need at least two face references to compose scenes; piping results into MiniMax H3 for video works perfectly, but they want a local starting-frame workflow to cut costs.
More from Multimodal
- Invideo launches an MCP for editing videos directly via Claude or ChatGPT — azed_ai · 2026-10-15
- DreamBooth creator Nataniel Ruiz leaves Google DeepMind, teasing a surprising new direction — natanielruizg · 2026-10-15
- AI-generated series 'The Engagement' drops episode 5 Friday, finale chapters in November — AIandDesign · 2026-10-15
- Seedance 2.5 'Would You Survive This?' Demo With Full Prompt Shared — umesh_ai · 2026-10-15
- Magnific co-produced AI documentary "Houdini: Back from the Dead" premieres on FOX Nation — cuenca · 2026-10-15
- Higgsfield drops Episode 2 of its AI influencer short drama — CurieuxExplorer · 2026-10-15