FlowEdit: Inversion-Free Text-Based Editing Using Pre-Trained Flow Models
[Paper] | [Project Page] | [Code]
🎨 Edit your image using Flow models! upload an image, add a description of it, and specify the edits you want to make
Notes:
- We use FLUX.1 dev and SD3 for the demo. The models are large and may take a while to load.
- We recommend 1024x1024 images for the best results. If the input images are too large, there may be out-of-memory errors.
- Default hyperparameters for each model used in the paper are provided as examples. Feel free to experiment with them as well.
Model Type
1 50
1 50
1 30
1 30
Examples
| Source Image | Model Type | Total Steps | src_guidance_scale | tar_guidance_scale | n_max | Source Prompt | Target Prompt | n_min (for improved style edits) | n_avg (improve structure at the cost of runtime) | seed | Output |
|---|