Smooth bottom overlay hotkey-release animation#222
Conversation
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 1d88130442
ℹ️ About Codex in GitHub
Codex has been enabled to automatically 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".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
| guard !self.layout.usesFixedCanvas else { return } | ||
| guard measuredHeight > 0 else { return } | ||
| self.dynamicPreviewMeasuredHeight = measuredHeight |
There was a problem hiding this comment.
Because BottomOverlayView is hosted once and reused across recordings, dynamicPreviewMeasuredHeight persists between sessions. This handler ignores the 0 height emitted when show() clears the preview, so a previous multi-line measurement can survive into the next quick/short dictation. When release mode starts, isBottomOverlayReleaseTransitioning then freezes the preview to that stale height, leaving a large blank transcription area visible during the processing animation until the overlay hides.
Useful? React with 👍 / 👎.
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 1dd746cdf0
ℹ️ About Codex in GitHub
Codex has been enabled to automatically 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".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
| let newlineCount = trimmed.filter { $0 == "\n" }.count | ||
| let charCapacityPerLine = max(Int((self.previewMaxWidth / max(self.layout.transFontSize * 0.56, 1)).rounded(.down)), 12) | ||
| let estimatedWrappedLines = max(1, (trimmed.count + charCapacityPerLine - 1) / charCapacityPerLine) | ||
| let maxVisibleLines = max(Int((self.previewMaxHeight / max(self.estimatedPreviewLineHeight, 1)).rounded(.down)), 1) | ||
| return min(max(estimatedWrappedLines + newlineCount, 1), maxVisibleLines) |
There was a problem hiding this comment.
In medium mode, the preview is laid out in a fixed-width ScrollView with a proportional font, so actual wrapping depends on glyph widths and explicit line breaks, not just trimmed.count. This new bucket calculation drives whether refreshSizeForContent() runs, so two previews can keep the same bucket while their measured height changes (for example, narrow text reflowing into wider words). When that happens the reused bottom overlay keeps its old height, leaving the last line clipped or stale blank space until some unrelated state change forces another resize.
Useful? React with 👍 / 👎.
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 57996f4de7
ℹ️ About Codex in GitHub
Codex has been enabled to automatically 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".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
| let hadLivePreviewText = self.asr.partialTranscription | ||
| .trimmingCharacters(in: .whitespacesAndNewlines) | ||
| .isEmpty == false | ||
|
|
||
| // Give SwiftUI a chance to render the processing state before we do heavier work | ||
| // (ASR finalization + optional AI post-processing). | ||
| await Task.yield() | ||
| // Only show a processing transition when we already observed spoken text. | ||
| // If there was no spoken text, let the overlay disappear immediately on hotkey release. | ||
| if hadLivePreviewText { |
There was a problem hiding this comment.
hadLivePreviewText is only true when a streaming partial has already arrived, but ASRService.start() only produces those updates for supportsStreaming models and very short utterances can end before the first chunk completes. In those cases this new guard skips menuBarManager.setProcessing(true) before await asr.stop(), so MenuBarManager.handleOverlayState() hides the overlay as soon as isRunning flips to false and users lose all Transcribing…/Refining… feedback during the slowest part of the flow. This regresses Whisper Medium/Large and short dictations compared with the previous always-visible processing state.
Useful? React with 👍 / 👎.
1 participant
Description
Smoothen the animation of the bottom overlay
Type of Change
Testing
swiftlint --strict --config .swiftlint.yml Sourcesswiftformat --config .swiftformat SourcesScreenshots / Video
https://drive.google.com/file/d/1BLZ0QM7V-ETxc6Otd5CbAPLWtx3GSBm6/view?usp=sharing