Local speech engine for developers. Run Whisper on your machine — ~44× faster than realtime. No cloud, no data leaves your device.
Install from source, add your audio file, run. All processing stays on your machine.
python -m venv .venv
.\.venv\Scripts\Activate.ps1
python -m pip install -r requirements.txt
.\run_gui.ps1python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
python -m guiFrom repo root. Full steps: Docs → Getting started.
Full quickstart →Local-first speech intelligence engine for developers and power users. The intelligence layer behind everything we build.
Transform any audio into structured, speaker-labeled text — completely offline. Dual-backend engine with automatic Commercial terms in preparation. 99+ languages.
Local-first vs cloud. Representative results; run your own tests with your hardware and data.
| Backend | Device | Avg RTF | Vs realtime | Offline |
|---|---|---|---|---|
| faster-whisper | GPU | 0.0226 | ~44× faster | ✓ Yes |
| faster-whisper | CPU | 0.0483 | ~21× faster | ✓ Yes |
| whisper | GPU | 0.0784 | ~13× faster | ✓ Yes |
| whisper | CPU | 0.1023 | ~10× faster | ✓ Yes |
Source: repo benchmark report (docs/product/BENCHMARK_REPORT.md). Tiny model, 4 files × 2 backends × 2 devices. RTF = runtime ÷ audio duration; lower is faster. Fastest single run: 0.0167 RTF (~60×).
$ python -m gui
GUI started. Open an audio file to transcribe.One-click presets tuned for specific industries and use cases.
Medical terminology, patient notes, consultation summaries.
Depositions, hearings, legal transcripts with precision formatting.
Action items, minutes, participant tracking, follow-up extraction.
Subtitles, captions, podcast transcripts with timing precision.
WhisperCore is the engine. Here's what we're building on top.
The speech intelligence engine. Local transcription, diarization, and export.
Call intelligence. Real-time transcription and insights from voice calls.
Meeting analytics. Automated minutes, action items, and participant insights.
Speech data insights. Trends, patterns, and intelligence across all your audio.
Core processing runs on your machine. No mandatory cloud upload. No usage-based billing. No surprise costs when volume spikes.
One license agreement for everyone — individuals and companies. Same rules. You pay per device per year. You own the compute, we license the software.
No mandatory cloud upload. Processing stays on your device.
Pay per device per year. Not per minute. Not per word.
One agreement for everyone. Companies scale by buying device licenses.
Choose file. Set options. Run. Open output. That's it.
Planned features — not yet released. From transcription tool to speech intelligence platform.
Auto-generated summaries from long transcriptions. Key points extracted instantly.
Automatically detect and extract action items, deadlines, and follow-ups.
Talk-time analysis, speaker roles, and participation metrics per speaker.
Automatic topic detection and keyword extraction from your transcripts.
Search across all your transcriptions by meaning, not just keywords.
We are finalizing plan tiers while preparing for beta.
WhisperCore is in pre-beta hardening.
Will be announced when beta readiness is reached.
Pricing is intentionally not finalized yet.
Packaging and licensing terms are in progress.
Offline and deployment requirements reviewed case-by-case.
Pricing: coming soon.
Runs fully offline.
Your data never leaves your machine.
Status FAQ - pre-beta access, offline use, and roadmap status.
Join the pre-beta waitlist. Pricing will be announced at beta launch.
Get beta updates