Krea 2 Identity Edit LoRA can place labeled objects exactly where you want
Volundai · reddit · 2026-07-27
A Reddit user shows how Krea 2 Identity Edit LoRA can be steered with annotated input images.
What the demo demonstrates
- If objects in the input image are marked with text labels, the LoRA seems to treat those labels as part of the prompt.
- The model then places the referenced objects where the user wants them in the generated scene.
Why it matters
The attached image shows a before/after example: a schematic input with labeled items like a window, sky, mountains, a blue toy robot, a red apple, and a large vase of flowers, paired with an output image that follows those placements in a realistic still life.
The post is essentially a playful but practical demo of controllable image editing and layout guidance.
More from Multimodal
- Third-party test: Claude Opus 5.5 renders finer 3D scenes but costs 13x more than GPT-6 Sol — testingcatalog · 2026-09-23
- ComfyUI trick: aux preprocessor + Qwen transfers poses across characters with one prompt — Acceptable-Work8202 · 2026-09-23
- Same portrait prompt across Midjourney V6.1, V7 and V8.2: do older models look better? — tisch_eins · 2026-09-23
- Testing AI character consistency across a 20-image travel sequence — SiennaVaire · 2026-09-23
- Midjourney v8.2 Faces: New Portrait Generation Samples Shared — azed_ai · 2026-09-23
- One-sentence prompt generates lifelike dog video, shown side-by-side with the real one — wgrathwohl · 2026-09-23