The 2.8.0 upload was refused with "description is too long": naming
Image Folder and Callout took the description to 1024 bytes and the
Registry takes 1000. It is 989 now, and release_preflight.py fails on a
description over the limit, so this is caught before a version lands.
The Registry's scan also warned about two statements on one line in
tests/test_image_folder.py and says that becomes an error. They are two
lines now.
A new Utility node for explaining a workflow on the canvas. The text grows
to the largest size that fits the note, so it reads without zooming in, and
each point gets a thin teal line beside it.
An arrow emoji in the text (or the arrow row under the editor) becomes a
teal arrow in the line. Click a node while editing, then press an arrow, and
the arrow is linked to that node: clicking it later brings the node into
view, the way Workflow Switches does. The link is kept in the text as #id,
so editing the words never moves it.
The node has one hidden text input and no outputs, so it never runs.
With format set to gif or webp the file was saved, but the node's stage
stayed empty and said "Execution finished without video metadata".
The node has shown a gif or webp in an <img> since the two formats were
added, because a <video> cannot play either. The run's result was still
searched with a list of video file endings only, so for these two nothing
was found and the <img> was never filled. The list of picture endings now
sits beside that lookup in js/shared/video_preview.mjs, and both the
lookup and the choice of tag read it. A png is still not taken.
In the classic renderer a saved webp was also drawn a second time under
the stage. ComfyUI shows an animated webp in a widget of its own, neither
its video preview nor its picture preview, and only those two were stood
down. suppressCoreVideoPreview hides that widget as well. The node's
smallest height with a saved webp goes from 469 px back to 415 px.
Tests: the lookup finds a saved gif and webp and sends them to the <img>;
every row of the backend's VIDEO_FORMATS is read from its source and has
to be found and sent to the right tag; the animated widget is hidden on
arrival and when it is already on the node.
In Nodes 2.0, Load Video and Save Video could not be dragged shorter than
their clip drawn at the node's full width. With a 9:16 clip, Load Video at
320 px wide stopped at 965 px, where the classic renderer goes to 584 px,
and Save Video at 823 px against 446 px. A node saved at its classic size
was drawn taller than saved and could cover the node under it.
Nodes 2.0 lays a node out by its content, and the saved size is only a
minimum. The shared stage kept its <video> in normal flow at full width, so
a portrait clip set the stage's height. The clip now sits out of flow and
fills the stage, and so does the <img> Save Video keeps for a gif or webp.
The stage takes its size from the node alone and the clip scales to fit.
The player fills the stage, where the preview picture is centred at its
own size, so its controls stay along the stage's bottom edge.
The classic renderer is unchanged.
A new node opens at its smallest size plus one more row of tiles
(340 x 602). That size was set a moment after the node was made, on any
node that sat at its smallest height. A workflow that loads has put the
saved size on the node by then, so a node saved at its smallest size
looked new and was grown: saved 320 x 506 came back 340 x 602, and saved
400 x 506 came back 400 x 602. A copy, a paste, an undo and Recreate node
grew the node the same way. A node saved 2 px taller kept its size.
The opening size is now set as the node joins the graph. The card is in
place by then, and a size put on the node afterwards wins: a loaded
workflow, an undo, Recreate node. A copy gets its size before it joins
the graph, so a node that is configured with a size is marked and left
alone. A node saved under its smallest size is lifted to it, and one
saved with no size opens like a new node, as before.
tests/image_folder.test.mjs fails when the opening size is set anywhere
but as the node joins the graph, or when a node that comes with a size is
no longer marked.
In Nodes 2.0 a node with a picture panel could be dragged shorter than the
panel's minimum. A corner drag took Mask Refine from 356 to 218 px and left
its picture a 2 px line, Save Image lost its saved picture, and Latent Size
lost its whole stage. The classic renderer stops at the panel's floor.
The classic floor is the minHeight a panel declares through fillNodeHeight.
Nodes 2.0 lays a node out by its content and does not use that number, so a
panel whose content has no height of its own had nothing to stop at.
fillNodeHeight now also writes the floor, less the DOM widget frame, to the
panel's element as a CSS min-height while the element sits inside a Nodes
2.0 node. It is written whenever the floor is read and when the panel is
first laid out, so a floor that follows state is followed: with the preview
switched off the node still shrinks to the bar.
The preview panel's stage also carries its own floor inside a Nodes 2.0
node. That panel's declared floor holds less frame than the shared number
takes off, and without the rule its nodes stopped 10 px under the height
ComfyUI lifts them to on the next load.
The classic renderer is unchanged.
A folder dropped on the node, or chosen in the file explorer from Browse or
Add, is copied into ComfyUI's input folder by the browser through ComfyUI's
own upload, and the node moves to the copy. A folder with more than 300
pictures or more than 1 GB asks first. Bringing the same folder in again
writes over its copies.
A whole path pasted into Folder is read as the folder it names when it lies
inside ComfyUI's input or output folder. Any other path is still refused,
and the node now says what to do instead, with a button that opens the file
explorer. The server reads nothing outside those two folders.
In Nodes 2.0 the node showed its source picture twice: on its pad canvas,
and again underneath as ComfyUI's own copy of the node's images. The copy
took the node's spare height: a node 900 px tall gave the pad canvas
278 px, where the classic renderer gives it 512. With a picture wired in,
the copy showed the file named in Source, which the node was not using.
The node's own suppression removed the classic preview widget and blocked
node.imgs, but never set hideOutputImages, the flag Nodes 2.0 reads before
it draws a node's images under the widgets. It sets the flag now. The
classic renderer never reads it.
tests/core_preview.test.mjs scans every frontend file that stands the
classic preview down by hand and fails when one does not stand the
Nodes 2.0 copy down as well.
Both are cards alone on their node. In Nodes 2.0 they stretched with a
node made taller, with empty space inside the card. They call
pinVuePanelHeight like the other cards now. Run Timer stays the one panel
that fills its node on purpose, and the guard names only it.
A loader for many pictures. The card chooses the folder inside ComfyUI's
input or output folder, and a thumbnail gallery under it ticks the pictures
to use. All in one run sends every picked picture through as a list; one
per run loads one picture each run and moves on when a run is queued. It
gives image, mask, filename, index and count. Pictures dropped on the node
or added with Add are uploaded into the input folder.
Folder names and picked names are attacker-controlled: a folder is only
read as a path below the input or output folder, links that lead out are
refused, and picked names are compared with what the folder holds, never
opened.
In Nodes 2.0 a node with a card above a picture panel gave half of its
spare height to the card. The card got empty space under its last row and
the picture was drawn at about half the size it has in the classic
renderer, on a node at its smallest size too.
Nodes 2.0 lets a widget's row take spare height whenever the widget has a
computeLayoutSize, and every DOM widget inherits one. A card that pinned
only its computeSize was pinned for the classic renderer alone.
pinVuePanelHeight clears the inherited method, and the widget cards and
Save Image's card call it. The classic renderer is unchanged.
Tooltips, two labels and one console warning still described older
versions of the nodes. Wording only: no inputs, outputs or saved values
change.
- Latent Size: the gear, its menu and an error said "Resolution".
- Video Crop + Rotate + Pad -> Clip: tooltips said Snap and Limit where
the node says Frames for and Length, and described dragging a handle
to move the whole window. IN and OUT trim; Length sets a frame count.
- Run Timer: earlier runs are in the right-click menu, not under the
number.
- Krea 2 Outpaint Model Patch: the rounding strip the warning names only
comes from Divisible by on the Crop + Rotate + Pad nodes. Load Image +
Pad never adds one.
- Save Image: its preview switch said off skips a temp file. It writes
none, so it gets its own tooltip; the other preview nodes keep theirs.
- Workflow Note: the Banner layout shows the title, subtitle and author.
- LoRA Loader: a docstring said the pack only reads the Civitai file. It
also writes it on Fetch Civitai info.
A width+height box with a 0 in it takes its shape from the source, so only rounding and Multiple separate the two. Pad answered that with a bar 1 or 2 px wide: a 1000x700 picture with Multiple 16 came out 1008x704 with a 2 px bar. The picture now fills the box. The output size is unchanged, and a box with both sides set pads as before.
Picking a frame writes IN. With the frame at or after OUT the node held a start after its end, and the next run stopped with "Load Video needs start_seconds smaller than end_seconds" until a handle was moved.
The card read a row's hover text from the widget, where the frontend does
not keep it, so no row of any card said anything on hover. It now falls
back to the node's definition, which holds the tooltips the backend
declares.
A card whose height changed set its node to its smallest size, so a node
made taller for its picture lost that height whenever More opened or
closed. The node now grows or shrinks by what the card did when the person
changed it by hand, stays as it is when the card changed by itself, and a
loading workflow takes its saved height. A node that sat at its floor
still hugs the new floor.
Fill holes fills gaps that lie fully inside the mask, with the same fill
Mask Refine uses. Free VRAM takes SAM 3 and its text reader out of the
graphics card's memory right after the search, with ComfyUI's own unload
of one model, also when the search found nothing. Both are off by
default.
Registers the node, gives it a widget card and the shared preview panel
with a FIND tool that runs the node alone, a readout of how many things
matched and how sure SAM 3 was, and the stop message on the node's own
face. A stopped run shows the plain picture, not the last find. Each name
asks SAM 3 for up to 16 matches, and Sureness starts at 0.7.
The card took any whole number with a range over 2048 for a pixel size and
stepped it by 8. On the Integer node that made a Steps box go 25, 17, 9, 1:
8 was out of reach without typing. Select Every Nth, Split Batch, Select
Frame and Load Video's frame limit stepped by 8 too.
A whole number now moves by its own increment. A row steps by 8 only when it
is a pixel size (a px unit, or `pixels` on a row that shows no unit, as
Image Resize's W x H), and a row can name its own `step` (Load Image + Pad's
Multiple keeps 8).
Every node stops at the width it was built for when it is dragged narrower:
the growing panels and the fixed cards. Load Video's IN and OUT boxes stay
whole in a narrow node.
The frontend adds 104 pixels to a growing panel's minimum width and leaves a
fixed-height card's minimum out of its width sum, so a card fell to a 210
pixel default. Panels pass exactMinWidth to fillNodeHeight or call
holdNodeMinWidth, and tests/panel_guards.test.mjs fails a new panel that does
neither.
Qwen Image 2.1's VAE decodes RGBA, so wiring its VAE Decode straight into
Stitch Inpaint failed with "Inpainted channels (4) do not match the cropped
image (3)". A 4-channel patch over a 3-channel stitcher now loses its alpha;
every other channel mismatch still raises.
A LoRA from the note's list that sits in a subfolder only showed as
installed, while the model, text encoder and VAE lines offered Use it. The
LoRA line now shows the folder and offers Use it as well, which writes the
found path into the LoRA Loader row. LoRA Loader follows a row change made
by another node in Nodes 2.0 too.
Use it also wrote the found path with forward slashes. A Windows server
lists files in subfolders with backslashes, and a loader only takes a file
exactly as listed, so the loader kept an unknown value. Use it now writes
the folder's own spelling.
2.6.2 let app shortcuts through from the boxes on a node face. Six more
boxes still stopped every key: the LoRA picker's search box, the trigger
word and template name boxes, the gear menu's boxes, the picture and video
picker's filter, and the server file path box on the video nodes. They use
keepKeyInField now, and a search box no longer treats Ctrl + Enter as a pick.
A test reads every keydown handler under js/ and fails on one that stops a
key before looking at it. The Workflow Note editor and the keep or discard
question are the two that own the keyboard on purpose.
Every field on a node face stopped all key presses, so ComfyUI's shortcuts did nothing while the cursor was in a Text node, a number box, the Seed box or a file name box. Ctrl + Enter only queued a run after a click outside the node.
The fields now keep the keys they type and edit with (so Delete or Ctrl + A never reach the canvas) and pass app shortcuts on to ComfyUI, as its own text boxes do. One helper, keepKeyInField in js/shared/canvas_passthrough.mjs, is used by the shared card's text, multiline and colour fields, the number boxes, the Seed box, Save Image's name fields, Load Video's IN and OUT boxes and LoRA Loader's strength boxes.
Updating the pack in Manager restarts ComfyUI and reloads the workflow, but
not the page. The tab kept running the old AusBoss scripts and nothing said
so: the update guard only looked once, when the page loaded.
The guard now looks again every time the server comes back, asks once per
installed version, and its message stays until it is closed instead of
going away after 30 seconds.
Follow-up to issue #108. The missing LoRA Loader panel reported there was
old browser copies of the shared modules, fixed in 2.5.2.
In Nodes 2.0 a panel sits in the node's own widget area instead of a
.dom-widget, so the pass-through did not recognise it and the drag-zoom
never started there. Both layouts are matched now.
The pack's server still never contacts a host. The info card's button has the
browser ask civitai.com about one file by its SHA256, then the server saves
the answer beside the LoRA as <model>.civitai.info after checking it names the
same file (size, shape and hash). The old Python HTTP call stays gone.
release_preflight now allows a raw fetch() only in js/cache_guard and
js/lora_loader. SECURITY.md, AGENTS.md, the README and the LoRA Loader help
page describe the change.
The wheel, Ctrl+Shift drag-zoom and middle-button pan never reached the
graph from over Compare, the crop previews, the video viewers or any control
on a card. One pack-wide listener hands them back, the way ComfyUI does for
its own text boxes. Panels that scroll keep the wheel until they reach an end.
Wired straight into a Qwen Image 2.1 sampler, the node's latent made a
picture twice as wide and tall as the size shown (1024 x 768 came out
2048 x 1536). The latent now carries how far it is downsampled, as
ComfyUI's own empty-latent nodes do, so ComfyUI fits it to a model that
packs pixels differently. Other models are unchanged.
The MaskEditor saves its picture into the image widget's stored value
without calling the widget back, so the node kept showing the old picture
and nobody could tell a mask was there. The stage now follows a source
changed that way (on the node's draw, and when the MaskEditor sets the
node's preview, which covers Nodes 2.0 too), keeping the framing.
The mask is the picture's alpha. The stage reads it once per picture by the
run's own 90% rule and draws the picture as the run uses it: kept pixels
solid, painted parts empty over the fill, with Show blend's teal on top.
The teal turns, crops and pads with the picture, in the full editor too,
and Show blend's paste mask now includes the painted parts. A gear on the
image source box turns the teal off for this browser.
Four orange corner handles sit outside the canvas corners, on the same
ring as the side diamonds. Dragging one makes the canvas bigger or
smaller and keeps its shape: the padding on that corner's two sides
changes and the opposite corner holds still. Alt (Option on a Mac)
changes all four sides about the centre, and can be pressed or let go
mid-drag. A corner stops when one of its sides runs out of padding.
Two Centre buttons, on the node face next to Fill and Feather and in the
editor's Padding section, put the picture in the middle side to side or
top to bottom. They only move padding from one side to the other, so the
canvas keeps its size and a lit ratio stays lit; with Divisible by on,
its strip counts, so the visible bands come out even.
The rotate knob took the spot the top-right corner handle now needs, so
it moves to the nearest free gap on the handle ring when blocked, and it
keeps its place on its arm while it turns. In the editor, Alt-drag still
pans everywhere except on a corner handle.
Frontend only: the node inputs are unchanged.
ausboss