GitHub Download

Model guide

Download the right model for your hardware.

These models power OSWispa. Bigger models sound better but demand more RAM and compute. Use the guidance below to avoid overload on modest systems.

Download models See hardware guidance
Resource snapshot CPU-only note

Rule of thumb

Base and medium models run on modest CPUs. Distil-large-v3 and large-v3 feel best with a GPU.

Actual usage depends on audio length, backend, and system load.

Disk space

142 MB - 2.9 GB

Model size on disk.

Suggested RAM

4 GB - 16 GB+

CPU-only baseline.

Speed range

Fast to Slow

Depends on hardware.

Downloads

Pick the model that fits your day.

Drop model files into ~/.local/share/oswispa/models/. Click a model to download.

Model Size Speed Best for
base.en 142 MB Fast Quick dictation
medium.en 1.5 GB Balanced Everyday writing
distil-large-v3 1.5 GB Fast Best all-around
large-v3 2.9 GB Slow Highest accuracy

Guidance

Model resource guidance for modest systems.

Suggested RAM is for CPU-only systems. GPUs can reduce latency significantly.

base.en

Smallest English model. Fast and light.

  • Download size: 142 MB
  • Suggested RAM (CPU-only): 4 GB+
  • CPU-only speed: good
Download base.en

medium.en

Balanced accuracy with heavier compute.

  • Download size: 1.5 GB
  • Suggested RAM (CPU-only): 8 GB+
  • CPU-only speed: moderate
Download medium.en

distil-large-v3

Fast for its accuracy. GPU recommended.

  • Download size: 1.5 GB
  • Suggested RAM (CPU-only): 8-12 GB
  • CPU-only speed: moderate
Download distil-large-v3

large-v3

Highest accuracy with heavy compute.

  • Download size: 2.9 GB
  • Suggested RAM (CPU-only): 16 GB+
  • CPU-only speed: slow
Download large-v3

Modest system tips

  • Start with base.en if you have 4-8 GB RAM or an older CPU.
  • Keep recordings short to reduce peak memory and latency.
  • Close heavy apps when testing medium or larger models on CPU-only machines.
  • Use distil-large-v3 or large-v3 only if your system stays responsive.