Neural ExpeditionSubscribe
TEST RECORD · ← all field reports

FIELD REPORTS/IMAGE/ISSUE #60

Edit photos with prompt instructions and generate fast images

Boogu Image is a practical image pick when you want one workflow that can handle both generation and editing. The most useful angle is simple: start in the browser, try a plain instruction on a source image, then decide whether the local setup is worth the extra control.

MODELBoogu Image
PUBLISHEDJune 23, 2026
READ TIME2 min
TESTED BYNeural Expedition
CATEGORYIMAGE

Field notes

01What it does

Boogu Image is an open image generation and editing family with Base, Turbo, and Edit variants. For this issue, the useful workflow is Boogu-Image-0.1-Edit: give it one reference image and a written instruction, then ask for a targeted change instead of generating a new image from zero.

That makes it easy to test on practical edits. You might upload a product shot and ask it to change the background, adjust the material, restyle the image, or replace part of the scene. The same release family also includes a faster Turbo path for text-to-image drafts, so the Space is useful as a quick way to compare prompt generation and image editing without installing anything first.

02How to try it

Start with the public Boogu Image Space. Use the Edit workflow first, not a polished gallery prompt. Upload one clear source image and give a narrow instruction such as changing the style, replacing the background, or modifying one object. On the first run, look for three things: whether the main subject stays recognizable, whether the edit follows the instruction, and whether text or small details drift.

For local testing, use the Boogu-Image-0.1-Edit model page and the project install notes. The local route is CUDA-oriented and expects the model weights plus the Boogu codebase. There is also ComfyUI support, which is likely the more practical path if you already test image models through node workflows.

03Caveat

This is still a research release, not a polished production tool. Local use needs a capable CUDA setup, the Edit workflow currently centers on one reference image, and the model card calls out instability around strict subject preservation, dense text, small faces, and complex body structure.

04What you can do with it

  • Edit an existing product or campaign image without starting from a blank prompt.
  • Test style changes, background changes, and object edits on the same source image.
  • Compare the Edit workflow with the faster Turbo generation path.
  • Prototype bilingual poster or product visuals with image and text constraints.
  • Build a local image-editing workflow around public weights and ComfyUI.

Try the demo

View model page