Krea 2 users share a tuning checklist for identity edit fidelity and likeness drift
vuse2121 · reddit · 2026-07-24
Krea 2 Identity Edit users discuss how to stop over-copying and identity drift
A Reddit user asks for a reliable troubleshooting order in Krea 2 Identity Edit v1.2 after repeatedly hitting two problems:
- The output becomes an exact copy of the reference image.
- The subject loses likeness depending on which reference image is used.
The post lists the suggested controls and how they affect results:
- refboost is the main fidelity dial: around 4 gives strong face/body likeness; above 10 can break removals; below 1 allows more creativity.
- Aspect ratio matters less on v1.2 fit geometry, but older weights and crop (legacy) need matched source/output AR.
- Keep generation at ≤ 2 MP; for two-person edits on v1.1, 1–1.5 MP is preferred to avoid identity blending.
- Step count is a mild tradeoff: fewer steps favor composition, more steps favor face detail; around 10 is recommended.
- groundingpx strongly affects adherence and likeness; too high can cause duplicate/split compositions.
- If CFG is above 1, the negative prompt should also be grounded.
- LoRA strength is suggested at 1.0.
It reads like a practical tuning checklist for users trying to stabilize Krea identity edits.
More from Apps
- Reddit users debate how much personal financial data to share with ChatGPT — sanbaeva · 2026-07-27
- AI Tool Reshapes Influencer Marketing: 150 Creators Filtered in 15 Minutes — yangyi · 2026-07-27
- Reddit users share how ChatGPT helps with anxiety and daily life — Jaded-Channel-7169 · 2026-07-27
- Open-source browser tool removes GPT Image 2 texture artifacts with a 0.48M model — Parking_Baby_57 · 2026-07-27
- Sentient OS brings proactive computer-use agents to the Mac with an on-device LLM — TechExpert2910 · 2026-07-27
- ChatGPT Sites turns a truck-service company’s 20-page inventory sheet into a mobile tool — Hacebeanbreakfast · 2026-07-27