Transcription Models & Cloud
Choose between local models and cloud APIs for transcription.
QuickContract supports two transcription modes: local (using Whisper models that run entirely on your Mac) and cloud (using OpenAI or Deepgram APIs for higher accuracy). You can switch between them at any time from Settings → Meetings.
Local models
QuickContract uses whisper-cpp, a high-performance C++ implementation of OpenAI's Whisper speech recognition model, to transcribe meeting audio locally on your Mac. The whisper-cpp binary runs as a sidecar process alongside QuickContract — it launches automatically when a recording starts and shuts down when the recording ends.
The Base model is bundled with QuickContract. You can download larger models from Settings for better accuracy.
Model comparison
| Model | Size | RAM needed | Accuracy | Best for |
|---|---|---|---|---|
| Base (bundled) | ~148 MB | ~1 GB | Good | Day-to-day English meetings, 8 GB Macs |
| Turbo | ~1.6 GB | ~6 GB | Very Good | 16 GB+ Macs, near-Large accuracy at 8x the speed |
| Large v3 | ~3.1 GB | ~10 GB | Excellent | 24 GB+ Macs, critical transcriptions, heavy multilingual use |
The Base model processes 10-second audio chunks faster than real time on Apple Silicon Macs. Larger models are more accurate but use more memory. QuickContract detects your Mac's RAM and shows a recommendation in Settings.
Stick with the Base model for local transcription, or enable Cloud Transcription for excellent accuracy with zero memory overhead.
Cloud transcription
For the highest accuracy without using local resources, you can connect a cloud transcription provider. Audio chunks are sent to the cloud API during recording and transcribed server-side. QuickContract supports two providers:
| Provider | Model | Cost | Notes |
|---|---|---|---|
| OpenAI | Whisper large-v3 | ~$0.006/min | Same model as the local Large v3, but runs on OpenAI's servers. Get a key from platform.openai.com. |
| Deepgram | Nova-3 | ~$0.0043/min | Purpose-built speech model. Excellent for accents, fast turnaround. Get a key from console.deepgram.com. |
When cloud transcription is active, QuickContract falls back to your local Whisper model automatically if the cloud provider is unreachable (e.g., no internet). Your local model is always available as a safety net.
When using cloud transcription, audio chunks are sent to the selected provider's API during recording. This is a direct connection between your Mac and the provider using your API key — QuickContract does not proxy or store audio on its servers. Review your provider's terms for their data handling policies.
Downloading a local model
Open Settings
Click the gear icon in the sidebar or press Cmd + , to open Settings.
Go to Meetings
Select the Meetings tab in Settings.
Download your preferred model
Under Whisper Model, you'll see the available models with download buttons. Click Download next to the model you want. The download progress is shown inline. Once complete, the model is stored locally and ready to use.
Select the active model
If you have multiple models downloaded, select which one to use for future recordings. You can switch between them at any time — the change takes effect on the next recording.
How transcription works
When a recording starts, QuickContract either launches the whisper-cpp sidecar (local mode) or prepares the cloud connection. Audio from both the system and microphone streams is processed in 10-second chunks. Each chunk is transcribed independently and appended to the running transcript with its speaker label ("me" or "them") and timestamp.
Because each chunk is small and self-contained, a transcription error in one segment doesn't affect subsequent ones. If a chunk contains silence or unintelligible audio, the transcriber simply returns an empty result and moves on.
Storage and management
Downloaded models are stored in QuickContract's application support directory. You can delete a model from the Settings panel to free up disk space. If you delete the active model, QuickContract will prompt you to download one before your next recording.