Providers & Models
Local mode: offline by default
Dimmy ships with Whisper running on your device. No API key, no internet, no per-minute cost. Works on a plane.
On first launch Dimmy downloads a Whisper model (around 78 MB for base-q8_0). After that, all transcription happens on your machine. The model uses your CPU or GPU; on Apple Silicon it runs through Metal, on NVIDIA it can use CUDA, on Vulkan-capable Windows GPUs it accelerates too.
What runs locally
- Speech-to-text via whisper.cpp.
- Filler removal (no model, just regex).
- LLM enhancement, optional, if you enable the on-device model. See Optional on-device LLM.
- Everything else: history, settings, key storage.
Performance
On a recent Mac, local Whisper transcribes a 10-second clip in roughly 0.6 seconds. On a 4-year-old laptop, around 1.5 to 2 seconds. With GPU acceleration, you get sub-second results for short clips on most hardware.
image · placeholder
When to use cloud instead
- You record long sessions (10+ minutes) and want them done in seconds.
- Your device is old or under-powered.
- You want absolute top-tier accuracy on noisy audio. Cloud Whisper Large v3 still has the edge.
Was this helpful?
In this category
Providers & Models
Cloud providers compared
Groq is fastest, OpenAI is the gold standard, Deepgram leads on streaming. All optional. Bring your own key.
Providers & Models
Add your API key
Open Settings → Providers & keys, find the provider card, paste the key. It is encrypted on your machine. You can update keys at any time.
Providers & Models
Choose a Whisper model
Smaller is faster but less accurate. base-q8_0 is the default. large-v3-turbo is the top tier on-device.
Providers & Models
Optional on-device LLM
Run the LLM enhancement pass locally via llama.cpp. No cloud, no key, no per-token cost. Pick a model to match your GPU memory.
Last updated: 2026-05-22