Krea 2 depth LoRA nails portraits, but action poses need literal prompts
iiTzMYUNG · reddit · 2026-07-23
A Reddit post shares first impressions of Krea 2’s depth LoRA, showing how well it recreates pose, camera angle, and composition from a depth map.
The author says it works very well for portraits and simple standing poses, but struggles with dynamic actions such as crouching or weapons. Their main advice is to write prompts in highly literal, stage-direction style so the model does not have to guess hand placement, limb direction, or head orientation. They also suggest using higher strength values, around 1.0–1.2, for tighter pose matching.
Related event: Krea 2 ControlNet Depth Model Tested(4 posts)→
More from coding & agent
- OpenAI ships GPT-6.1 Terra; Brockman says its computer use now beats his own speed — gdb · 2026-10-27
- Kolab Benchmark Uses Pro Screen Recordings to Show AI Agents Overestimate Collaboration Quality — minnesotanlp · 2026-10-27
- Dev argues Claude burns far more tokens than GPT models for the same coding work — haider1 · 2026-10-27
- Stop building agents that can't change models: stable contracts, external state, real evals — rseroter · 2026-10-27
- Cheap models can cost more in agent harnesses than frontier ones, researcher warns — vasuman · 2026-10-27
- Kanzo: A 5k-Line Rust Kanban Binary With MCP for Claude Code and Codex — allmudi · 2026-10-27