Skip to content

FreeVideo 0.4.17 - #389

Merged
KBRASK merged 17 commits into
mainfrom
release/0.4.17-public
Oct 20, 2026
Merged

KBRASK merged 17 commits into
mainfrom
release/0.4.17-public

Conversation

@KBRASK

@KBRASK KBRASK commented Oct 20, 2026

Copy link
Copy Markdown
Collaborator

0.4.17: Image generation and editing come to Apple silicon Macs, a Mac update that fails to open is rolled back, and phones on the same network connect more reliably.

  • Mac: Macs with Apple silicon and at least 16 GB of memory can now generate and edit images.
  • Mac: in later updates, if the new version stops right after opening, the previous version is put back. The previous version is kept in a hidden folder so it no longer shows up in search or Launchpad. When an update fails, the reason and both buttons now appear right below the description.
  • Phone access: the phone page connects when phone access is switched on in the launcher, and also works on older phone browsers.
  • Queue: a single failed read of the queue no longer shows the connection lost card. The card appears after three failures in a row and goes away when the connection returns, and a running loop is kept through a short outage.
  • Launcher: when the chosen folder is not a ComfyUI folder, the launcher names a ComfyUI folder found nearby. Windows: the engine now starts when its usual port falls inside a range that Windows has reserved.
  • Installation: an installation with no generation on record now also compiles the standard size in the background.
  • Failures: when the model file changed during generation, the failure text names what changed.

KBRASK added 17 commits October 19, 2026 03:33
…ght and Launchpad never...

Mac: keep the old app in a .noindex folder during an update so Spotlight and Launchpad never show a second FreeVideo
… pass on an installation...

Background kernel compile also covers the Medium level's larger first pass on an installation with no request on record
…y wording

An image attention call that MLX cannot run now goes to the Torch attention
instead of failing the picture, and later calls in the same process skip MLX.

The message for a Mac below 16 GB now states the fact first. The character
card's failure note on a Mac also says memory rather than GPU memory, as the
picture editor already does.
When the folder chosen for ComfyUI does not contain it, the launcher now
looks up to three levels above and one level below for a ComfyUI folder and
names it, and shows a dedicated card asking for the folder that contains
main.py or the portable folder that contains it, instead of the generic
failure heading.
The retained error carries a traceback whose source lines quote the message,
so the folder name was never read from the exception line itself. Read it
from the exception line, show one instruction instead of two, and test with
the retained traceback.
The phone page combined its request signals with AbortSignal.any and
AbortSignal.timeout, which iOS Safari before 17.4 and Firefox before 124
do not have. Every request threw before it was sent, so the page showed
the connection-lost banner at once. The signals are now joined by hand
with the same 10 second limit, and a browser test removes both functions.
The launcher saves its LAN switch in its own folder, but the agent routes
read the copy next to the installation, so every request the access proxy
forwarded for the phone page was refused as local only. Access now needs
all copies of either place to say yes and the most recently saved copy to
say yes, so a switch turned off in one place stays off.
A saved installation that can no longer be found ends with the same sentence
as a wrong folder choice, but the user chose no folder. Leave that error on
its own path and test it with the retained traceback in both languages.
A streamed checkpoint identity failure now adds one line to the failure
text that says which comparison and which fixed fields differed (never
paths or timestamp values), so feedback reports show whether the change
was to the modification time, change time, size or file identity.
… a loop stopped by a long one

A single dropped queue read left the lost-connection card on screen, stopped a
running loop and was never cleared. The queue now shows the notice only after
three polls in a row fail, adds how long the outage lasted and whether the
browser was online to the error details, keeps a loop running through the
outage, and removes the notice when a read succeeds again.

While the connection stays down the polls slow down to at most one every eight
seconds. A running loop that has seen reads fail for three minutes is stopped,
and its notice is built again with the final length of the outage and the
reason. Further failed reads only extend the outage, so the stopped-loop notice
stays until the user acts. Without a loop the notice still clears on the next
successful read. Other read errors and the browser-extension path behave as
before.
Text to image, edits, reference pictures and region edits now run on the
built-in image engine on Apple GPUs. Weights stay views of the mapped model
files and are lent to the GPU one layer at a time, the int8 layers use the
Apple GPU int8 kernels with a dequantizing fallback, and long attention runs
through MLX's fused kernel.

A Mac plans against half of its memory, which sets the size tiers it offers.
A Mac below 16 GB is told it needs 16 GB or more instead of failing later,
and out-of-memory notes on a Mac speak of memory rather than the graphics
card. CUDA, ROCm and CPU paths are unchanged.
@KBRASK
KBRASK merged commit 58ccf44 into main Oct 20, 2026
5 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant