How Processing Works

Processing runs in the cloud after your upload finishes. You can follow the stage on the session page or close the tab and come back later.

Video recordings

1. Transcription starts in the background as soon as the audio is extracted.
2. Candidate frames are sampled every few seconds and at scene cuts. Longer recordings are sampled at longer intervals.
3. Filtering drops frames that are too dark or too bright.
4. Slide detection compares frames, keeps one image per slide, and skips returns to a slide that was already shown. Recordings from the macOS app use settings that tolerate cursor movement and overlays.
5. Slide text is read and passed to the AI, so the notes use the terms on the slide.
6. Slide images are re-captured from the video at up to 2560 pixels wide.
7. Alignment matches the transcript to the slide that was on screen.
8. Notes are written: a short key message and up to three bullet points per slide, plus a summary, key takeaways, action items, and a session title.
9. The report is written next. It walks through the recording section by section, with the slides as figures.

Long recordings are split into parts of about 45 minutes and processed in parallel. If a recording has no audible sound, LensNote makes visual notes without a transcript. If no slides are found, the recording is treated as an audio session.

Audio recordings

The recording is transcribed with speaker labels, then LensNote writes a summary with three to six key takeaways, any action items, a session title, and a one-page summary image.

Language

LensNote detects the spoken language from the start of the recording and keeps it for the whole session. The notes, summary, and report are written in that language. The summary image is written in English, or in Chinese for Chinese recordings. There is no language setting.

Speaker labels

Transcripts can include labels such as Speaker 1 and Speaker 2. Labels are assigned to short stretches of audio, so on longer recordings the same person can appear under more than one label. Check names before you quote someone.

How long it takes

It depends on the length of the recording. Recordings of 30 minutes or longer send you a "Your video notes are ready" email when they finish, unless you have turned that email off in notifications.

Accuracy

The notes are generated by AI from the transcript and the slides. The AI is instructed not to add facts that are not in the recording, but it can still make mistakes. Check important details against the recording: click a slide or a transcript line to play that moment.

If processing fails

The session shows "Your recording is safe" and a Retry Processing button. A retry continues from the last completed step where possible. Minutes reserved for a failed or cancelled job are returned to your balance. LensNote does not send an email when a job fails.