Importing meetings

Not every meeting starts with Abschrift's record button. Drop a Zoom or Meet recording, an exported transcript, or notes from another tool anywhere over the app window — or click the small arrow next to the record button. Imports become normal meetings: notes, speaker labels, search, MCP, and the same retention schedule as everything else. Transcription and speaker separation run on this Mac.

What you can drop

• A recording — audio or video. Audio formats include wav, m4a, mp3, aac, and flac. Video formats include mp4, mov, and webm. Abschrift extracts the audio track, transcribes it on-device, and separates speakers as "Speaker 1", "Speaker 2", and so on.
• A transcript — VTT, SRT, or a text or Markdown export. Zoom's and Google Meet's export formats are recognized by default. That includes Meet's block timestamps and Zoom's wall-clock times, normalized to meeting time.
• Notes — a text or Markdown document, imported word for word and marked as externally written. Headings, links, lists, tables, quotes, and code render in the app in Abschrift's restrained style.
• Any combination, one meeting per drop. With a recording and a transcript, the transcript is used as-is by default. To run local transcription and speaker separation instead, choose "Re-transcribe on this Mac" in the confirm dialog.

Before anything is written, a confirm dialog shows exactly what the import will use. It lists each file's detected type, which you can correct. It previews the first converted turns and prefills editable title and date-&-time fields.

The title comes from the transcript's content when it names one, otherwise from the file name. The date works the same way, with the file's modified date as fallback. Adjust either, or clear them. An empty date means "the moment of import". An empty title falls back to the file name. If a turn has no usable time in the source, it keeps no timestamp — Abschrift never invents one.

On confirm, the meeting appears in the sidebar immediately and processes in the background, with the same progress view as a recorded meeting. Nothing blocks: you can keep using the app.

Afterwards, re-run any step from the meeting's ⋯ menu whenever its input still exists: Re-generate notes uses the transcript, Re-transcribe uses kept audio. Images embedded in an imported notes file render in place. Web-hosted images load only when you click them (privacy policy).

Unusual formats

Standard formats convert entirely on-device. When a transcript's format isn't recognized, Abschrift asks your configured notes provider to describe the format. The provider receives a short sample — at most the first 60 lines — and returns a conversion rule. The transcript text itself is always taken verbatim from your file. A fidelity check rejects any rule that doesn't explain at least 95% of the file's lines.

With Ollama this detection runs fully locally. With a cloud provider, the sample is sent only after you confirm the import, and the meeting's egress badge records it. Learned formats are remembered, so each format needs detection at most once. Without a configured provider, unrecognized formats fail with a plain message. Convert to VTT or SRT and try again.

Retention, storage, and honesty

• The imported transcript gets a retention stamp at import time and is deleted on schedule like any other. The original dropped file stays untouched where it was. The copy kept with the meeting is deleted together with the transcript.
• Imported transcripts and notes are marked as imported in their file headers. They are the external tool's output, converted, and never claimed as Abschrift's own transcription.
• Imported meetings count toward the free tier's monthly budget like recordings, by the day you import them.
• Renaming speakers and reassigning segments work on imported transcripts. Voice recognition across meetings only applies when you re-transcribe a recording locally. Text imports never touch the voice-profile store.

One thing importing does not change: consent rules for recording a meeting apply wherever it was recorded. See Recording & the law.

Command line

The same import runs headlessly: abschrift import call.m4a transcript.vtt. Add --dry-run to see the detection verdict as JSON without importing.