Built for

Interviews, transcribed
on your terms.

Every interview ends as a speaker-labeled transcript in your project folder. The audio is processed by a transcription account you signed up for and control — not by a sub-processor we chose and you have to get approved.

Download for macOS Apple Silicon · 14-day trial · $29 launch

The approval problem

Interview work runs into the same wall everywhere: your consent form, your ethics review, or your client's DPA names who may process participant data. A note-taking service that joins the call and stores recordings in its own cloud adds a party you now have to document, justify, and sometimes get contractually approved — for a tool you only wanted for transcripts.

Heliotype removes itself from that chain. It records locally and has no server. When you transcribe, the extracted audio goes to your AssemblyAI account under your key — you hold that contract and their DPA directly, and can name them in your paperwork as the one processor involved.

To be exact, because this is the part people get burned on: transcription is still cloud transcription, and AssemblyAI is still a third party — just your third party. We make no compliance claims and hold no certifications. Check the arrangement against your IRB, DPO, or client DPA before you rely on it. A fully on-device engine is the top item on the roadmap, with no date promised.

What the workflow gives you

  1. Speaker labels and timestamps — [0:04:12] Speaker A: — which is the format that makes coding and quoting possible in the first place.
  2. Named by session. Recordings take their name from the calendar event, so P07Interview-20260728-1400.m4a matches your participant log instead of a timestamp you have to decode.
  3. Plain text beside the audio. Both files land in the folder you chose, ready for Dovetail, Atlas.ti, NVivo, a repo, or a plain grep.
  4. Nothing joins the call, so a nervous participant is not staring at a third-party bot in the participant list while answering your questions.
  5. Existing recordings too. Right-click any audio or video file in Finder and choose "Transcribe with Heliotype" — useful for the archive of sessions you already ran.

Cost, at research volumes

Transcription bills to your key at roughly $0.17–0.23 per audio-hour, and a new AssemblyAI account includes about $50 of credit — on the order of 200 hours before you pay anything. A 20-participant study with hour-long sessions fits inside that credit. The app itself is $29 once, which for once means a grant line rather than a recurring subscription you have to defend annually. Run your own numbers.

What it will not do for you

It does not summarize, tag, or theme your interviews — deliberately. It hands you clean text and stops; the analysis stays in your tools and your head (or the model you drag the transcript into). There is no live captioning during the session, no Windows version, and it needs an Apple Silicon Mac. Accuracy also degrades with crosstalk and heavy background noise, as every automated transcript does — plan to spot-check quotes you intend to publish. The full list is here. Related: institutions that block bots, client work, and the full FAQ.