Download and Manage Models

This guide is for anyone setting up TypingVoi models for recording, file transcription, or local cleanup. It shows how to use Model Library to install the right models, understand what each model is for, and avoid confusion when multiple profiles share the same downloads.

1. Open Model Library

Open TypingVoi, then open Settings and select Model Library from the sidebar. This page is the shared home for model downloads, install status, and model storage across the app.

Screenshot placeholder: Model Library page in Settings with the sidebar selection visible and both the STT Models and LLM Models sections on screen

The page is split into two main sections:

• STT Models for speech-to-text during recording and file transcription.
• LLM Models for local transcript cleanup.

If you only need voice-to-text, start with an STT model. If you also want AI cleanup to run locally on your Mac, install a local LLM as well.

2. Understand what each model type does

Before installing anything, it helps to match the model type to the job:

• STT models turn speech into text. These are the models TypingVoi uses when you record from the menu bar or transcribe an audio file.
• Local LLM models rewrite or clean up transcripts after transcription. They do not replace the speech-to-text model.

Model Library is shared across profiles. That means one installed model can be reused by multiple profiles instead of being downloaded again.

3. Install a model

In either section, find the model you want and click Install. TypingVoi downloads the model locally and shows progress in the same row.

Screenshot placeholder: A model row showing the Install button before download and a second row state with Installing percentage and progress bar

Use this workflow:

1. Open Settings > Model Library.
2. Decide whether you need an STT Model or a Local LLM.
3. Find the model row you want.
4. Click Install.
5. Wait for the row to finish downloading and return to its installed state.

A few practical rules:

• Install at least one STT model before you rely on recording or file transcription.
• Install a local LLM only if you plan to use transcript cleanup locally.
• You do not need to install the same model more than once for different profiles.

4. Refresh the library when models look missing

Use Refresh Library in the STT Models card if the catalog looks incomplete or the app does not show the models you expect yet.

Refreshing is useful when:

• you just opened Settings for the first time
• you finished onboarding and want the latest model list
• local LLM options are not appearing yet
• a model was expected to be installed but the page looks out of date

Refreshing updates the shared catalog. It does not mean you have to reinstall models that are already present.

5. Use the language filters to narrow the list

Model Library includes two separate filters:

• Language filter in STT Models narrows speech-to-text models by spoken language support
• Text language filter in LLM Models narrows local cleanup models by the text languages they support

Use the filters when the full list is longer than you need or when you are setting up a profile for a specific language. If you are unsure, leave the filter on All and start with the model TypingVoi recommends elsewhere in onboarding or profile settings.

6. Watch the size and shared storage impact

The Model Library page shows storage awareness in two ways:

• individual model rows can show an estimated size
• the page header shows the total installed model storage

This matters because models are shared resources. Installing several large models can use noticeable disk space even if you only actively use one or two of them.

A good default is:

• keep one or two STT models that match your real languages and workflow
• add a local LLM only if you will actually use cleanup on-device
• avoid collecting extra models just to compare them later unless you need to test accuracy

7. Use the “Used by” line to understand profile relationships

When a model is already assigned in one or more profiles, Model Library shows a Used by line in that model row. This tells you which profiles currently depend on that shared model.

That helps when:

• you want to know whether a model is safe to stop using
• you are standardizing several profiles on the same model
• you are troubleshooting why one profile still points to an older model

If you want to change which model a profile uses, do that from the profile’s language and model settings. Model Library tells you the relationship, but profile selection still happens in the profile itself.

8. Decide what to install first

If you want the fastest path to a working setup, install in this order:

1. One STT model for your main spoken language.
2. Test recording from the menu bar.
3. Add a second STT model only if you truly need another language or a different speed and accuracy tradeoff.
4. Add a local LLM later if you decide to use transcript cleanup on-device.

That keeps storage predictable and makes it easier to tell which model is affecting your results.

Related guides

• Record From the Menu Bar and Use Push to Talk
• Transcribe Audio Files on Mac
• Choose Languages, Models, and Custom Vocabulary
• Clean Up Transcripts With AI

Best next step: install one STT model, then move to Record From the Menu Bar and Use Push to Talk to test a real dictation workflow.