Skip to content

Fix/UI field corner radius - #9

Merged
pdjdev merged 52 commits into
build/osxfrom
fix/ui-field-corner-radius
Oct 30, 2026
Merged

pdjdev merged 52 commits into
build/osxfrom
fix/ui-field-corner-radius

Conversation

@pdjdev

@pdjdev pdjdev commented Oct 30, 2026

Copy link
Copy Markdown
Owner

No description provided.

ivandrofly and others added 30 commits October 28, 2026 21:03
Replace INamesList.Load with Task LoadAsync. SeNamesList now reads the
name XML files with XmlReader's async API (Async = true settings,
ReadAsync / ReadElementContentAsStringAsync) instead of blocking.

FixCommonErrorsViewModel awaits the load in RunScanWithFeedbackAsync,
the delete handler and the skip-step-1 startup path rather than inside
the synchronous ApplyFixes. BatchConverter.FixCommonErrors becomes
FixCommonErrorsAsync. Test fakes, SeNamesList tests and the benchmark
setup follow the new signature.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Applies the review from @GitGianluc on SubtitleEdit#16447: 'pen' in YouTube Timed Text
means a set of formatting options, not a physical pen. Also rewrites the
karaoke hint and fixes 'Right' translated as 'Giusto' (correct) instead
of 'Destra'.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…tt-pen-strings

Italian: YTT pen strings as formatting options, fix Right
- SeNamesList.LoadAsync parses the name files with the sync XmlReader on a
  worker thread into fresh sets and swaps them in when done, so lists in use
  are never seen half-filled; a generation counter keeps a slower earlier
  load from overwriting a later one.
- Fix common errors: concurrent callers await the running load instead of
  starting another, a load failure is logged and shown instead of being
  lost in the fire-and-forget startup path, and skip-step-1 startup goes
  through the guarded "Analyzing..." scan.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
All 14 download URLs and 28 hash keys (archives + unpacked executables)
move from v0.8.46 to v0.8.48. Archive hashes match the release API
digests; inner folder names and bundled CUDA DLLs are unchanged, and
the CLI flags and backend list are identical to v0.8.46.

v0.8.48 keeps word boundaries when Parakeet emits standalone boundary
markers (CrispStrobe/CrispASR#558).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Without VAD, crispasr runs Granite in fixed 30 s auto-chunks that drop
words at the cuts. On FLEURS dev (30 en / 25 de clips, peak-normalized,
crispasr 0.8.48, SE's default --max-len 50 --split-on-punct), Silero
VAD took Granite 4.1 from 11.8% to 9.2% WER in English and from 12.7%
to 10.7% in German, and was slightly faster. Granite 5.0 improved the
same way.

Granite now joins Cohere, Mega and Index-Echo in the "Auto" VAD set.
--vad, --vad-model or --chunk-seconds in the advanced parameters still
take precedence, and the empty-result retry still runs without VAD.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…e-force-vad

Use Silero VAD for Crisp ASR Granite by default
Two new shortcuts (no default key): "Move start X ms back, push previous
end" and "Move end X ms forward, push next start". They move by the same
"Move start/end shortcut step (ms)" setting as the plain X ms shortcuts,
but on reaching the minimum gap they push the neighbour's border along
instead of stopping. Unlike the KeepGap frame variants, the neighbour is
never pulled back when moving away again. Touching lines stay touching,
and the pushed line is not squeezed below the minimum duration.

Requested on doom9 for keyboard-only timing (Alt+Left/Right).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…ms-push

Add "move start/end X ms" shortcuts that push the neighbour
…al-cue-drag

Fix read-only original cues moving during waveform dragging
The timeline's original row drew each row's effective original time codes,
which follow the translation while it is retimed (lock on, or a line that
shares its row's times). A read-only original never moves, so draw its
lines' own time codes instead - the same fix SubtitleEdit#16465 made for the waveform.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…ne-readonly-original-drift

Keep read-only original blocks still in the timeline while retiming
…#16469)

While a relative step's seek is still in flight, the player reports the old
position, exactly one step behind the tracked target. The flat 0.5 s resync
guard then dropped the previous step: at the limit for 500 ms steps (float
rounding) and always for longer steps. Widen the guard to 0.5 s + step size;
every other position change already clears the tracker.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…leedit-issue-16469-d7d087

Fix uneven 500 ms / custom video steps on quick presses (SubtitleEdit#16469)
…e match (SubtitleEdit#16471)

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…leedit-issue-16471-3afd21

Word lists: arrow keys continue from the type-ahead match (SubtitleEdit#16471)
…leEdit#16472)

AddName indexed every name - plain ones too - in the dash/period part
lookup, and RemoveName never took it out, so HasNameExtended kept the
word as a name and Undo skipped straight past it. Only names with '.'
or '-' are indexed now (as at load), and RemoveName drops them.

NameList.Remove also blacklisted every removed name, and Add refused
blacklisted names, so a name added again after Undo was never saved.
Only shipped names are blacklisted now, and an explicit Add clears the
blacklist entry.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…leedit-issue-16472-e60927

Spell check: Undo of "Add to names list" shows the word again (SubtitleEdit#16472)
Updated the Japanese translation for SE5.4.0 Beta10
ffmpeg's subtitles filter ignores the [Fonts] attachment section, and the
styles preview did not carry the footer anyway, so a style using an
attached font previewed in a fallback font. The libass screenshot now
decodes attached fonts to a temp folder passed as fontsdir, and the styles
preview passes the subtitle footer (re-rendering when attachments change).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…mbedded-font-preview-752542

ASSA styles: preview attached [Fonts] fonts
The SSA styles preview draws with Skia, which only finds installed fonts by
name, so a style using an attached font previewed in a fallback font. The
attachments are now decoded to Skia typefaces (cached until the footer
changes) and the face matching the style's font name is passed to
TextToImageGenerator via a new optional typeface parameter.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…bedded-font-preview

SSA styles: preview attached [Fonts] fonts
niksedk and others added 22 commits October 29, 2026 20:52
Picking a format in the format picker (right-click the format combo, or the
shortcut) went through SetSubtitleFormat, which marks the change as
programmatic. That skipped the format conversion and left Save as on the
default Save as format, so picking Sub Station Alpha suggested SubRip and .srt
instead of Sub Station Alpha and .ssa.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The macOS save panel appends the preferred extension of the format's file
type when the suggested name has none, and .ssa shares a type with .ass
(io.mpv.ass), so Save as Sub Station Alpha suggested movie.ass.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…ation-alpha-default-0475dc

Save as Sub Station Alpha: picked format and .ssa extension
The "Second subtitle file" dialog always used the video preview font.
Add a Font name row (drop-down + "..." font picker). It is saved with
"Remember these settings"; stored empty while it is the preview font so
it keeps following that font (SubtitleEdit#13492).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…le-properties-font-picker-9af497

Second subtitle: add a font name picker
A second file already timed from zero is no longer shifted by the
video offset; one with the same offset times as the main file still is.
Each reading is scored by how many lines start inside the video (or the
main subtitle span), tie keeps the offset. Work on SubtitleEdit#16434

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…etect-offset-abd3bf

Second subtitle: auto-detect if the video offset applies
The "Edit rule" section under the rules list is gone. Enter, F2 or a
double-click now edit the selected rule in its own row (find, replace
with, description - the caret goes in the clicked cell); Enter keeps
the change, Escape puts the texts back, leaving the row keeps it, and
the preview follows while typing. The type icon in front of "Find what"
is now a button that picks the rule type. "Edit..." still opens the
full dialog. The rules and preview panels lose a little padding.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…le-replace-compact-d2c10b

Multiple replace: edit rules in their row, drop the Edit rule panel
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…-release-prep

v5.4.0-beta11 release prep
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…ify, margin area, colors

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…-overlay-text-fixes

FFmpeg player overlay: closer to the mpv preview
…tings

Linux always installed the CPU build, so -ngl had no effect. The build
picker now also shows on Linux, and on every re-download from the engine
settings dialogs on Windows and Linux, so an installed build can be
changed. Automatic updates keep the installed build.

A switch now removes the previous build's GPU backend files, which
unpacking would otherwise leave behind. The Linux build is detected from
libggml-vulkan.so for the Backend label and updates.

Picker idea, Linux test and doc section from SMAP-git in SubtitleEdit#16488.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…cpp-linux-vulkan-16488

llama.cpp: pick CPU or Vulkan on Linux, switch builds from engine settings
@chatgpt-codex-connector

chatgpt-codex-connector Bot commented Oct 30, 2026 •

Copy link
Copy Markdown

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review Status Commit Review trigger
📝 Code Review ✅ Completed 2026-10-30T17:03:02.065787Z 0ae5c4a PR opened
ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

@pdjdev
pdjdev merged commit 25f016d into build/osx Oct 30, 2026
4 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.

6 participants