Choose your platform
Each button downloads the current installer, version 1.0.0. The builds are signed and update themselves after that.
On a Snapdragon or other Windows ARM laptop, take the native build instead: it is many times faster than running the x64 version under emulation. Windows on ARM64 (Snapdragon)
What you are downloading
| Current version | 1.0.0, released 15 October 2026 |
| Platforms | macOS (Apple Silicon and Intel), Windows x64, Windows ARM64 |
| Speech models | Whisper Small, Medium, Large v3 Turbo — pick one in Settings, downloaded on the spot |
| Spoken languages | Over 90, detected from the speech itself |
| Interface languages | Eleven, listed below |
| Exports | DOCX, PDF, TXT, Markdown |
| Installer | Signed, with a built-in updater |
System requirements
- macOS 12 or later — Apple Silicon (M1 to M4) and Intel
- Windows 10 (1903 or later) and Windows 11 — x64 and ARM64
- 8 GB RAM recommended; 16 GB is comfortable for Large v3 Turbo
- Optional GPU acceleration: NVIDIA RTX 20-series or later, AMD, Intel Arc
No discrete graphics card is required. Without one, transcription runs on the processor at roughly real time with the medium model, so a one-hour recording takes about an hour.
What happens on first run
Scrieb needs the internet once, to download the speech model you pick. After that, transcription, speaker labels, summaries and chat all run on your computer with no connection at all.
1. Install and sign in
Sign in with your email address and a magic link. The licence syncs across your devices, and an offline licence keeps recordings working without internet.
2. Download a model
Choose Small, Medium or Large v3 Turbo. You can record while it downloads; those recordings are transcribed as soon as it finishes.
3. Work offline
Drop in a file or record directly. Subscribers can switch on Offline mode, after which Scrieb makes no connection on its own at all.
Need a video tutorial? Here it is on YouTube
Supported files
| Import | MP3, WAV, M4A, AAC, FLAC, OGG, Opus, WMA, AIFF, 3GP, MP4, MOV, M4V, WebM, MKV |
| Export | DOCX, PDF, TXT, Markdown |
Video files are accepted and the audio track is transcribed. Drag files onto the window, or use Import to say what they are so the right summary type is used.
Interface languages
The app runs in English, German, French, Spanish, Portuguese, Russian, Dutch, Swedish, Danish, Norwegian and Finnish. The Windows installer appears in your Windows display language rather than always in English, and summaries are written in the language of the recording.
Free plan and pricing
The free plan covers one hour of recording a day, with speaker labels, summaries and every export format, and does not ask for a card. A paid plan removes the daily cap; there is no per-minute billing on any plan.
Pricing
One licence for Mac and Windows. No minute limits, speaker labels included, free to try.
Changelog
The three most recent releases. Updates install themselves, or you can check for them under Settings → General.
1.0.0 · 15 October 2026
- The spoken language is detected from the speech itself rather than the first 30 seconds, so a recording that opens with silence is no longer read as the wrong language.
- Silence and other non-speech passages are skipped, so Whisper stops inventing text over quiet stretches.
- A transcript that does not match the speech is marked "Check transcript" instead of being summarised.
- Calls recorded with "Record from App" through speakers no longer carry an echo, and the other person is no longer mistaken for you.
- Export to Word (.docx), keeping headings, bullets and quotes as real Word formatting.
0.8.1 · 12 October 2026
- Summary types: Interview, Lecture, Meeting, Voice memo, or your own instructions.
- Custom vocabulary, so names, product terms and acronyms are spelled your way.
- Library search looks inside transcripts and summaries, not just titles.
- AI chat can answer across all your recordings and names the ones it used.
- Danish, Norwegian and Finnish added, eleven interface languages in all.
- Offline mode for subscribers: no update checks, usage statistics or crash reports.
0.8.0 · 9 October 2026
- Native Windows ARM64 build for Snapdragon X and other Windows ARM laptops.
- Choose where transcription runs: automatic, a specific GPU backend, or the processor.
- Transcribe a recording again with a different model; the earlier transcript is kept.
- The computer no longer sleeps mid-transcription, which used to turn minutes into hours.
Frequently asked questions
- Is Scrieb free to download?
- Yes. The download and the free plan cost nothing and need no card. The free plan covers one hour of recording a day, with speaker labels, summaries and all export formats. A paid plan removes the daily cap.
- Do I need an internet connection?
- Once, to download the speech model. After that transcription, speaker labels, summaries and chat run entirely on your computer, and you can disconnect completely. Subscribers can switch on Offline mode, after which the app makes no connection on its own at all.
- Which Windows file do I need?
- If your PC has an Intel or AMD processor, take the standard Windows download. If it is a Snapdragon or another Windows on ARM laptop, take the ARM64 build: the x64 version runs there only under emulation, which makes transcription many times slower.
- Does it work on Intel Macs?
- Yes. The macOS build is universal and runs on both Apple Silicon and Intel, with macOS 12 or later. Apple Silicon is considerably faster, and Intel Macs are best paired with a smaller model.
- Is there a Linux version?
- No. Scrieb is available for macOS and Windows only. On Linux, the Whisper command-line tool and WhisperX are the usual route, though neither labels speakers.
- Which Whisper model does it use?
- You choose: Whisper Small, Medium or Large v3 Turbo, under Settings → AI Models. Larger models are more accurate with accents, technical vocabulary and background noise, but slower and heavier on memory. You can transcribe a recording again with a different model and keep both results.
- Do I need a graphics card?
- No. A GPU speeds things up and NVIDIA RTX 20-series or later, AMD and Intel Arc are all supported, but the processor alone works. Expect roughly real time with the medium model, so a one-hour recording takes about an hour.
- Is the installer signed?
- Yes, on both platforms, and the app verifies its own signature before applying an update. Updates install themselves, or you can check manually under Settings → General.