## Summary
- Fixes SaveWebM node showing "Error loading image" in Vue nodes mode
- Extracts `isAnimatedOutput`/`isVideoOutput` utility functions from
inline logic in `unsafeUpdatePreviews` so both the litegraph canvas
renderer and Vue nodes renderer can detect video output directly from
execution data
- Uses output-based detection in `imagePreviewStore.isImageOutputs` to
avoid applying image preview format conversion to video files
## Background
In Vue nodes mode, `nodeMedia` relied on `node.previewMediaType` to
determine if output is video. This property is only set via
`onDrawBackground` → `unsafeUpdatePreviews` in the litegraph canvas
path, which doesn't run in Vue nodes mode. This caused webm output to
render via `<img>` instead of `<video>`.
## Before
https://github.com/user-attachments/assets/36f8a033-0021-4351-8f82-d19e3faa80c2
## After
https://github.com/user-attachments/assets/6558d261-d70e-4968-9637-6c24532e23ac
## Test plan
- [x] `pnpm typecheck` passes
- [x] `pnpm lint` passes
- [x] `pnpm test:unit` passes (4500 tests)
┆Issue is synchronized with this [Notion
page](https://www.notion.so/PR-8943-fix-detect-video-output-from-data-in-Vue-nodes-mode-30a6d73d365081e98e91d6d1dcc88785)
by [Unito](https://www.unito.io)
---------
Co-authored-by: github-actions <github-actions@github.com>
## Summary
<!-- One sentence describing what changed and why. -->
Added feature to drag and drop multiple images into the UI and connect
them with a Batch Images node with tests to add convenience for users.
Only works with a group of images, mixing files not supported.
## Review Focus
<!-- Critical design decisions or edge cases that need attention -->
I've updated our usage of Litegraph.createNode, honestly, that method is
pretty bad, onNodeCreated option method doesn't even return the node
created. I think I will probably go check out their repo to do a PR over
there. Anyways, I made a createNode method to avoid race conditions when
creating nodes for the paste actions. Will allow us to better
programmatically create nodes that do not have workflows that also need
to be connected to other nodes.
<!-- If this PR fixes an issue, uncomment and update the line below -->
https://www.notion.so/comfy-org/Implement-Multi-image-drag-and-drop-to-canvas-2eb6d73d36508195ad8addfc4367db10
## Screenshots (if applicable)
https://github.com/user-attachments/assets/d4155807-56e2-4e39-8ab1-16eda90f6a53
┆Issue is synchronized with this [Notion
page](https://www.notion.so/PR-8282-Batch-Drag-Drop-Images-2f16d73d365081c1ab31ce9da47a7be5)
by [Unito](https://www.unito.io)
---------
Co-authored-by: Alexander Brown <drjkl@comfy.org>
Co-authored-by: Austin Mroz <austin@comfy.org>
Under a combination of many edge cases, the `widget_values` migration
code added in #3326 would cause the progress text on a "Recraft Text to
Image" node to incorrectly deserialize into the `control_after_generate`
- widgets_values is of length 1 greater than it should be because
progress text serializes
- It should not, there is no code to deserialize it
- negative_prompt has force_input set and skips serialization
- Migration only applies when `widgets_values` is equal to actual inputs
length. The two above edge cases cancel to make this true
- Seed is accounted for when calculating the length of widgets, but not
when applying the migration
- Migration occurs even though we track workflow version now and have an
accurate way of determining that it can not be needed
The two primary edge cases which cause the bug are both addressed
- `options.serialize` does nothing and has never done anything. I've
been guilty of making the same mistake in the ancient past, and want to
clean up the misconception where I can.
┆Issue is synchronized with this [Notion
page](https://www.notion.so/PR-8625-Fix-incorrect-widgetValue-migration-2fe6d73d365081a683b4c675eaeebb6c)
by [Unito](https://www.unito.io)
- Updated all imports from '@comfyorg/litegraph' to '@/lib/litegraph/src/'
- Replaced deep dist imports with direct source paths
- Updated CSS import in main.ts
- All imports now use the @ alias consistently
