Hands-on with Fibo 2.0 image editing: solid 360° turns, but object occlusion fails
Big_Parsnip_9053 · reddit · 2026-10-21
A Reddit user tested Fibo 2.0 image editing locally via a ComfyUI workflow (fibodevint8convrot diffusion model, qwen3vl4b text encoder, flux2-vae; 50 steps, CFG 5, Euler).
- 360° turntable test: rotating a character through successive angles while preserving the rest looked decent, with occasional eye/hand issues; the model didn't just copy poses and naturally kept the book in her hands when she turned away.
- Object permanence test: choosing an image with a hair ornament and deliberately not mentioning it, the ornament was pasted on even at angles where it should be occluded — no better than Qwen or Flux.
- Additional tests covered removing background characters/text and reconstructing the scene. Single-seed, uncurated outputs, so results may understate the model's ceiling.
More from Multimodal
- Same-seed image edit model comparison: Flux 3 Klein alone nails pose and jacket side — Dunc4n1d4h0 · 2026-10-21
- Fully AI-made short film 'The Life of the Gorgon': Grok writes, Kling 4.0 animates — nima_owji · 2026-10-21
- OmniCam from Tencent Hunyuan generates camera trajectories from panoramas and text — Zhenyang Liu · 2026-10-21
- Longtime Suno user roasts its vocal shimmer artifact: 'makes me puke' — lfschiavo · 2026-10-21
- Reddit user shares AI-made Death Note short film for TheLoopHole Project — neptune765 · 2026-10-21
- Minimax H3 Character Replacement: Ref2va 8-Step LoRA Transfers Facial Emotion Acting — mexxmillion · 2026-10-21