Models and add-ons

Not sure which model to pick? Read Choose a model first.

Download and switch

Pick a family (Sabich, Jachnun, Shakshuka, Parakeet) and you'll see its versions as cards. Each card has its own Download, Use this one and Delete, so you can download a model without switching to it. The model in use is outlined in green with an In use badge.

Help me choose… asks about your device and what matters to you, then recommends one.

Downloads are checked before use. Each model is on disk once.

Bundles

Bundles download a set in one click and skip anything you already have.

Bundle	Contains
Everything	Sabich, Shakshuka, and the clean-up, title and paragraph add-ons
Voice Typing	Sabich and Shakshuka
Transcription	Sabich with the clean-up, title and paragraph add-ons

Updates

Check for updates looks for new versions of every model and add-on you've downloaded, and Update all installs them in one go.

Hebrish also checks by itself. By default it installs new model versions automatically, and the old version is deleted once the new one has downloaded and passed its check. Change how often it looks, or turn off automatic installs, in Settings → Model updates.

Where models are stored

Where the models are stored shows the folder and opens it. Change folder… picks another, and the app offers to move your downloaded models there.

Add-ons

The Utilities tab has small models that work alongside the speech model, offline.

Add-on	What it does	Size
Silero VAD	Tells speech from fans, traffic and café noise. Included with the app	Included
Paragraph model	Finds where paragraphs start in a long transcript	34 MB
S1-mini by Superwhisper	Clean up: the default clean-up model	485 MB
S1-ASR-Cleanup	The same job with a fine-tune of S1-mini, experimental	662 MB
Title model	Writes a short title for saved transcripts. You set the longest title here (60 characters by default)	292 MB
Transform model	Qwen3 4B Instruct, for Transform	2.5 GB

Sizes are as shown by Ubuntu 1.025.