Local MiniMax H3 workflow matches Nano Banana image quality, user dropping Google Flow
Domskidan1987 · reddit · 2026-09-04
A Reddit user reports getting Nano Banana 2/Pro-level image quality locally by repurposing the MiniMax H3 Ref2V workflow: remove the save-video node, add a Get Image by Batch node (index 0 and 1) plus a save-image node, set megapixels to 3–5 and duration to 0, and use an LLM prompt rewriter. The same modification to FL2V and I2V workflows works surprisingly well for image editing — start prompts with 'A still frame shot of the last frame first…'.
Caveats: video-as-image output suffers VAE degradation past the first frame, and the 0.0 duration default outputs 4 frames; always grab frame 0 as it's the cleanest. The author says this setup finally lets him cancel his Google Flow accounts.
More from Multimodal
- Ant open-sources 5 Ming-3.0 speech models spanning TTS, audio understanding and ASR — Med1_Ai · 2026-10-21
- What keeps AI videos watchable: hooks, consistency and a payoff — aftahi_ai · 2026-10-21
- MORPHJUTSU skill in Flova transfers any video's motion to your character — aftahi_ai · 2026-10-21
- NVIDIA open-sources MotionBricks: 350K motion skills at 15,000 FPS, 2ms latency — aftahi_ai · 2026-10-21
- Runway joins Claude Marketplace, letting enterprises spend Anthropic budget on video generation — runwayml · 2026-10-21
- Kokon 0.10 turns Qwen-Image 2.1's text encoder into a prompt writer, no extra model — Ok-Director4174 · 2026-10-21