[Feature] Annotate gallery images with comments and add them to a chat draft

[achose369]

Problem / use case

When reviewing images or screenshots an agent shares in chat, I often want to point directly at an issue and explain what needs changing.

Currently, I have to take another screenshot, annotate it in an external tool such as Windows Snipping Tool, and paste it back into T3 Code. This interrupts the feedback workflow when the image is already available in the gallery.

Sometimes I need to explain changes across several images, so annotating one image should add it to a draft, not immediately send a message.

Proposed solution

Add a Draw / Annotate button to the image gallery, alongside the close button. In annotation mode, show basic drawing controls and an optional comment field below the image.

Provide an Add to message button that adds the annotated copy and its comment to the current chat draft without sending it.

Suggested workflow

1. Open an existing image in the chat gallery.
2. Click Draw / Annotate.
3. Draw arrows, circles, or strike-through marks to identify the relevant areas.
4. Add an optional comment explaining the requested changes.
5. Click Add to message. The annotated image and its comment enter the composer; nothing is sent.
6. Repeat for other images as needed.
7. Review the combined draft, add overall instructions, and send when ready.

Each image should remain clearly paired with its own comment so the agent knows which instructions apply to which image.

Example and mockup description

The proposed layout has:

• A pencil button in the upper-right corner of the gallery, beside the close button, to enter annotation mode.
• Colored marks drawn directly on the image to identify specific areas and requested changes.
• A comment field beneath the image with an Add to message action.

For example, on a screenshot of a settings panel, an accompanying comment could say:

Adjust the spacing where the green arrow points, add a close button at the yellow mark, and remove the row crossed out in orange.

The user could then annotate a second screenshot and add it to the same draft before sending both together.

Smallest useful scope

• Freehand drawing with color selection.
• Undo and clear annotations.
• An optional comment for each annotated image.
• Add to message, without automatically sending.
• Multiple annotated images and their comments in one draft.
• Ability to remove an image or revise its comment before sending.
• Preserve original images and any existing composer content.
• Support both agent-shared images and user-uploaded images.

Dedicated arrow, rectangle, and text-label tools could be added later.

Why this helps

This keeps visual feedback inside the conversation. Users can show exactly where a change is needed, explain it alongside the image, and collect feedback across several screenshots before asking the agent to act.

It also avoids re-capturing images, losing detail, or accidentally including unrelated surrounding UI.

Related request

#1429 proposed overlapping functionality and was closed as not planned.

I'd like to ask whether this focused gallery-to-agent feedback workflow could be reconsidered: annotate images already displayed in T3 Code, associate comments with them, and collect them into a message draft. The main purpose is to make visual instructions to the agent easier.

[achose369]

[AlexanderHott]

+1 for this, being able to comment on a region of an image available with t3code would be great

[SbokyZahodi]

Need it