From a live mark to a finding your agent can act on.

Four steps, one local session

1. Pick only what you mean to share

Choose a window, region, or display. Window mode stays locked to the selected native source.

2. Narrate and mark the moment

Speak naturally and hold Cmd on macOS or Ctrl on Windows while dragging to create a finding.

3. Stop and let the pipeline align it

Whisper transcribes, the aligner matches each timestamped mark to nearby words, and the selected provider structures the result.

4. Review or paste the report path

Edit titles and severity if needed, then paste the copied Markdown path into the coding agent or export another format.

The processing pipeline

Screen + Voice → Whisper → timestamped transcript → Aligner → marks matched to surrounding words → Provider → structure and severity → Generator → Markdown / HTML / JSON / PDF / tracker-ready

Graceful degradation is part of the design

With no ffmpeg, MarkuprPlus can still produce transcript-only output. With no Whisper model, timer-based screenshots preserve visual checkpoints. With no provider, Local Rules writes the report deterministically. processing-trace.json records the path actually taken.

Everything saved for one session

originplayer-test-iphone-20260818-104731/
├── feedback-report.md
├── feedback-summary.md
├── metadata.json
├── processing-trace.json
├── screenshots/
│   ├── marked-issue-001.png
│   ├── marked-issue-002.png
│   └── marked-issue-003.png
├── session-recording.webm
└── session-audio.webm

Default keyboard shortcuts