mirror of
https://github.com/open-webui/open-webui.git
synced 2026-05-06 19:08:59 -05:00
[GH-ISSUE #21806] issue: 0.8.5 image editing always uses first uploaded image id #19576
Reference in New Issue
Block a user
Delete Branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Originally created by @iChristGit on GitHub (Feb 23, 2026).
Original GitHub issue: https://github.com/open-webui/open-webui/issues/21806
Check Existing Issues
Installation Method
Git Clone
Open WebUI Version
0.8.6
Ollama Version (if applicable)
Latest
Operating System
Win
Browser (if applicable)
No response
Confirmation
README.md.Expected Behavior
On prior versions, there was some fixes that made it that each new image that been uploaded to chat can be edited by the LLM, but on latest version whenever user upload a second image to chat and ask for editing, the LLM references the old image that he edited before.
Actual Behavior
LLM always referenced the first image that has been already edited
Steps to Reproduce
Upload one image and ask for a change and then upload the second image and ask for a different change
Checked with many LLMs and with comfyui as the image backend
Logs & Screenshots
Additional Information
No response
@iChristGit commented on GitHub (Feb 23, 2026):
https://github.com/open-webui/open-webui/issues/20754
Older related issue, was working on 0.8.0
@Classic298 commented on GitHub (Feb 24, 2026):
how do you use these models? For the builtin image editing, it works fine for me.
@iChristGit commented on GitHub (Feb 25, 2026):
First uploaded image indeed get edited (Using native tool calling gpt:oss-20b and comfyui backend)
Second image user uploads do not go to the LLM, it keeps using the first uploaded image, user can just copy the ID of the new image and give it to LLM and it does work, so its just about passing the correct "image_url" id string.
@tjbck commented on GitHub (Feb 25, 2026):
We'll either need a more detailed instructions on how to reproduce, or a screenrecording, we're unable to reproduce.