- Rename test to reflect that it only validates the in-memory BytesIO stream-copy
path; documents that the .mp4 file-path case (SaveVideo's default) drops alpha.
- Flush the libvpx-vp9 decoder with decode(None) in the helper.
- Rework BriaTransparentVideoBackground description to make it explicit that this
is currently the only way to persist the alpha channel to disk.
- Move 'Transparent' to the end of the dropdown and set explicit default='Black' so
newly-added nodes preserve prior behavior (review warning).
- Reword node + tooltip descriptions to be honest that Save Video drops the alpha
plane when its default .mp4 extension is used; direct users to the IMAGE+MASK
node for any workflow that needs to persist alpha to disk today.
- Update the dispatch comment with the same caveat.
Bria's /remove_background endpoint accepts background_color='Transparent'
when paired with an alpha-capable container/codec; pass output_container_and_codec=
'webm_vp9' in that case (otherwise mp4_h264, unchanged). The returned WebM
flows through ComfyUI's VIDEO type as a file reference, so Save Video with the
default auto/auto settings stream-copies it byte-for-byte with the alpha plane
intact (covered by new test_save_to_auto_preserves_vp9_alpha_via_stream_copy).
BriaTransparentVideoBackground is kept as-is because it remains the supported
path for the IMAGE+MASK workflow: Get Video Components currently cannot expose
alpha for VP9 streams (PyAV's default vp9 decoder drops the side alpha layer),
so the per-frame compositing workflow still needs the dedicated node that
decodes via libvpx-vp9 itself.
* Move dataset/text nodes to text category
* Rename category utils into utilities
* Rename category api node into partner
* Move categories conditioning, latent, sampling, model_patches, training, etc. under model category
* Dispatch partner nodes in to 3d, audio, image, text, video categories
* Move PreviewAny node to utilities category
