How to directly edit an input image instead of generating one first?

[wangzrk]

Hi, thank you for the great work on this project!

I have a question regarding image editing functionality. Currently, it seems that the workflow requires generating an image first before any editing can be done. I was wondering if there is a way to directly edit a user-provided input image (e.g., uploaded by the user) without having to generate a new one beforehand?

If this feature is not currently supported, are there any plans to add it, or any recommended workarounds?

Thanks in advance!

[mi804]

https://github.com/modelscope/Nexus-Gen/blob/main/image_editing.py this code can directly edit an input image

[wangzrk]

Hi, thank you very much for your previous kind and prompt response — it helped me resolve the issue I was facing.

Now I'm trying to finetune the vision decoder on my own dataset. I noticed that in the finetuning code, the dataset seems to be built from both image and embedding files.

I would like to ask: how is the embedding obtained in this case? Is there a specific script or preprocessing step I should follow to generate these embeddings from my dataset?

Thanks again for your excellent work and support!

https://github.com/modelscope/Nexus-Gen/blob/main/image_editing.py this code can directly edit an input image

[mi804]

Please refer to #16 (comment).

[wangzrk]

Please refer to #16 (comment).

Thanks a lot!