Transcribe the room. Know who said what.

Transcription Studio is a native app from FCT Technologies that turns a recording into a transcript that knows who was speaking. It runs on macOS and iPhone from one codebase and does the entire job on your device. One model recognizes the speech and a second one tells the speakers apart, so a meeting, a call or a room recording becomes a labeled, timestamped who-said-what transcript. No transcription service is involved at any point. There is no cloud provider, no API key, and no third party that ever receives the audio. Storage is a separate question, and we answer it plainly rather than let the on-device claim imply it: the app requires an FCT account, and your library (sessions, transcripts and the recordings themselves) is stored in that account so it reaches your other devices. On the Mac it can also take a public URL or a dropped media file. On iPhone it works from files. It grew out of FCT's open-source transcription web engine, which still runs behind the company's own workflows and is published under the MIT license at github.com/fcttechnologies/TranscriptionStudio-Web, so the lineage is readable rather than asserted. The native app is built and in daily internal use. It is not currently sold or listed on the App Store.

Diarization is working out who spoke when. It is handled by NVIDIA's streaming Sortformer model, which tells up to four speakers apart and runs locally on the Apple Neural Engine. Getting it on the device took real work. The public Core ML version of Sortformer would not load on any current toolchain, so FCT re-exported the model from NVIDIA's original checkpoint with the current export tools. Because a speaker label is only useful if you can verify it, the app is built to be checked rather than believed. An Inspector shows the model's raw output (a per-frame speaker-activity heatmap, recognition confidence, and per-stage latency) and lets you click any segment to play its audio and judge the label by ear, or run the same session through a second, independent diarizer and compare the two timelines. Underneath, an automated test suite guards every build with a known-answer audio gate, a synthetic ground-truth attribution test, and a cross-check that fails loudly on disagreement.

Yes. Privacy is the whole design. Every stage runs on your own device: the speech model, the speaker model on the Apple Neural Engine, and the media handling through local tools. No outside provider is ever in the loop and there are no API keys, so the audio is never handed to a transcription service. Storage is the separate question, and the app is explicit about it. An FCT account is required, and your library (sessions, transcripts and the recordings themselves) is stored in it so it reaches your other devices. What matters for sensitive material is that FCT is the only party that ever holds it, and no transcription vendor is in the loop at all. That makes it a fit for client calls, interviews and internal meetings, where handing audio to a hosted service is not acceptable. When capturing a call's system audio on macOS, the app asks for the standard screen-recording permission, and you stay in control of what is captured. The open-source web engine that Transcription Studio grew from follows the same local-first principle and is published under the MIT license, so you can read exactly what it does.

Yes. Before the native app, Transcription Studio was a small, self-hosted web service, and that engine is still open source under the MIT license at github.com/fcttechnologies/TranscriptionStudio-Web. It runs Faster-Whisper locally behind a FastAPI backend, with yt-dlp and FFmpeg handling media download and preparation, and it is comfortable on a 16 GB Mac mini with no GPU. You paste a link or drop a file and get a clean transcript back, through the one-link web page or a JSON API you can automate against. It runs on macOS, Linux or Windows and you can self-host it in an afternoon. It still runs inside FCT's own operating system as the transcription layer behind every 'transcribe this' request.