Comparison
A recording preserves the session. MarkuprPlus extracts the findings.
Both formats have value. The difference is whether the recipient must watch and interpret the whole timeline or can act on one aligned block of evidence per problem.
What changes for the agent
| Question | Plain screen recording | MarkuprPlus |
|---|---|---|
| What the agent receives | A video file it may not be able to watch | Markdown with one annotated PNG per finding |
| Three problems in one session | One file with three scribbles to interpret | MX-001, MX-002, and MX-003 as separate findings |
| Narration | An unaligned audio track | A transcript segment aligned to each mark |
| Capture context | Usually none beyond visible pixels | Window ID, app, platform, cursor, and focus hints |
| Does the tool film itself? | Often, unless the operator hides it | No; picker, canvas, and HUD use OS-level content protection |
| Where transcription runs | Wherever the recipient chooses to process it | On-device Whisper by default |
| Output formats | Video and its audio track | Markdown, HTML, JSON, PDF, and tracker-ready templates |
When a plain recording is still better
Timing-dependent bugs can depend on a sequence that is clearer when watched end to end. Plain video is also better when the recipient is a human who would rather watch than scan a report, and for animation glitches and scroll jank where motion itself is the evidence.
MarkuprPlus does not throw that evidence away: the session video and audio are kept alongside the extracted findings for exactly those cases.
Timing-dependent bugs where order and latency matterHuman review where watching is faster than readingAnimation glitches and scroll jank