LoRA Training Help: Outputs Mismatch Sample Images
Kryimsson · reddit · 2026-08-03
A novice user is experiencing issues while training a character LoRA. They created a dataset using Image-to-Image (I2I) for accurate likeness and trained it on Civitai using the Krea2 model.
The Issue: The 3 sample pictures from the final epoch looked exactly like the target face. However, when the LoRA is put into action, the generated faces look nothing like the samples, regardless of adjusting the weight to 0.7, 0.8, 0.9, or even 1.0.
The author is asking the community for advice on specific settings or whether local training is necessary.
More from Multimodal
- Testing Seedance 2.5: A Seamless 30-Second One-Take Musical Instrument Transformation — LinusEkenstam · 2026-08-04
- ByteDance Launches Seedance 2.5: 30s Generation, 50 Multimodal Reference Assets — SucceededMind · 2026-08-04
- Ostris AI Toolkit Adds MiniMax H3 T2V and I2V Training Support — ostrisai · 2026-08-04
- Ostris AI Toolkit Adds LoRA Training Support for MiniMax H3 Video Model — ostrisai · 2026-08-04
- Multimodal Model Omni Accurately Renders Dynamic UI from Complex Prompts — fofrAI · 2026-08-04
- Runway Announces Seedance 2.5: Unlimited Generations for New Max Subscribers — runwayml · 2026-08-04