Help Center · AI and study

AI processing

When audio is in, SummitScribe runs a background pipeline. The website only enqueues the job; a worker does the work. You poll status — you do not wait on a single browser request.

Stages (in order)

1. Transcribing — speech-to-text on the audio (long files are chunked).
2. Speaker — names / labels for who spoke. If naming times out on a very long meeting, the job can continue with simple labels rather than failing everything.
3. Sectioning and notes — divide the meeting and write per-section notes.
4. Synthesis and vision — meeting-level takeaways, actions, summary; image analysis merged in. Pass the Test then merges per-section flashcards/facts and runs a practice-quiz pass (batched; length scales with the lecture).
5. Finalizing — mark the meeting complete and clear stale error flags.

Progress appears on Overview and on the AI Processing tab.

Statuses you may see

Partial is your recovery: the worker will not auto-retry it in a loop.

Reprocess

On AI Processing (and recovery actions when a job failed):

• Re-run from speakers
• Re-run sectioning / one section
• Synthesis only — reuse sections, cheaper (used when you switch to a template that was never processed)
• Generic retry if the stage name is cancelled or unknown

Reprocess uses processing time according to the work. Prefer the smallest action that fixes the gap.

Hours

The transcribed duration is what counts toward your monthly cap, recorded when transcribe completes. Deleting a meeting issues a reversing usage event so the meter stays honest. See Plans and billing.

If it looks stuck

1. Open the meeting AI Processing tab — read the canonical job, not a cancelled sibling.
2. Wait out a long speaker or notes stage (multi-hour meetings take minutes to tens of minutes).
3. If status is partial, use a targeted reprocess. Do not spam full reprocess.
4. Email info@summitscribe.app with the meeting title and time.