Fixing identity shift in I2V: Will training a WAN LoRA help?
__MichaelBluth__ · reddit · 2026-08-02
A user noticed severe identity loss during slight camera movements in Image-to-Video (I2V) generation, even when using a Krea2 character LoRA image as the first frame.
The author started a discussion asking if training a dedicated character LoRA for Wan2.2 would fix this identity shift, seeking practical advice on:
- Dataset requirements for training a WAN LoRA.
- The feasibility of reusing the Krea2 dataset and configuration settings.
- Recommended training configurations.
Related event: Tackling Character Inconsistency in Wan I2V via LoRA Training(2 posts)→
More from Multimodal
- ForcingBooster halves computation of autoregressive video diffusion with patch-wise denoising — VAST-AI · 2026-10-16
- Tokyo's Gemini AI Photo Booth turns visitors into plush toys and dioramas — bradneuberg · 2026-10-16
- Orchestrating a 100-image portrait of humanity for aliens with ChatGPT, Argon and p5.brush — dx8xb · 2026-10-16
- Indie Creator Builds an AI-Native Animated Show Solo — ComprehensiveLime654 · 2026-10-16
- Ideogram teases 4.5 arrival and open weights: 'Not yet, but soon' — East_Shoe8814 · 2026-10-16
- Opus 5.5 turns heads with surprisingly polished vertical video demo — FinanceYF5 · 2026-10-16