Fibo 2.0 LoRA Trained in 42 Minutes for $2 Puts Plush Dragon Into Famous Paintings
SirToki · reddit · 2026-10-21
A Redditor trained a character LoRA for Bria's Fibo 2.0 one day after release, inserting a plush dragon mascot into famous images like The Scream and Abbey Road.
- Config: official trainlora.py, 1024px, rank 64, alpha 64, lr 1e-4 constant, AdamW, batch 1, grad accum 2
- 400 steps (likeness at 200, best at 300); rented A100 40GB on Vast, 23GB VRAM peak, 6.2 s/step, 42 min total, $2 in GPU time
- Findings: works well with plain edit instructions; adding a dataset reference image improves details but often changes the whole picture; on Abbey Road it replaced a Beatle instead of adding a fifth person
More from Multimodal
- MiniMax H3 I2V: how to stop secondary faces' eyes drifting in 9:16 clips — Downtown-Term-5254 · 2026-10-21
- Mistral's Voxtral-mini-tts-3-0 debuts at #6 of 20 on Voice Arena Hindi TTS board — CodeByPoonam · 2026-10-21
- Offline CLIP image search for manufacturing defect photos: full pipeline walkthrough — OkShirt9372 · 2026-10-21
- Reddit asks for best open-weights image editing model for upscaling and reframing — subnohmal · 2026-10-21
- Wan 3.0 lands on Higgsfield at roughly 1/4.5 the price of Seedance 2.5 — sven_ai · 2026-10-21
- User seeks tools to turn a story into 40-120 coherent, framing-aware image prompts — Malina_Fox · 2026-10-21