Downloading and managing local models

Arqenne runs its AI features on your own Mac. To do that, it needs a few model files — the voice engines and the writing model. There is nothing to install by hand: these download automatically the first time you launch the app, and the first time you use a feature that needs a larger model. Once they are on disk, everything works offline.

This guide explains what gets downloaded, where it lives, how much disk to expect, and how to manage it.

What gets downloaded

Arqenne uses three sets of models, each powering a different feature:

Model set	Approx. size	Powers	Requirements
Standard Text-to-Speech	~90 MB	Fast, low-latency narration in the editor's Audio mode — an on-device neural voice engine plus a starter set of voices	Any supported Mac
Enhanced Text-to-Speech	~3.5 GB	Studio-quality narration and voice cloning	Apple Silicon (uses the GPU)
AI writing model	~0.5 GB (larger on Macs with 12 GB+ of memory)	AI writing suggestions in the editor	Apple Silicon (uses the GPU)

A few things worth knowing:

Standard Text-to-Speech downloads on first launch so basic narration is ready quickly. It is the lightest of the three.
Enhanced Text-to-Speech and the AI writing model are larger and download when you first reach for the feature that needs them, or shortly after launch. On Macs with 12 GB or more of memory, Arqenne automatically downloads a larger, higher-context version of the writing model for better suggestions; everything else works the same.
You do not choose a model or pick a file. Arqenne selects the right one for your Mac.

Models download from a model host the first time only. After that, generation happens entirely on your Mac — your text and audio never leave it.

Watching the download

The first launch needs a network connection so Arqenne can fetch the models. While a download is running, Arqenne shows progress in the app so you can see it working. Larger models (Enhanced Text-to-Speech especially) take longer on a slow connection, so it is normal for that one to keep downloading in the background while the app is already usable for other things.

You can keep working while models download. Standard Text-to-Speech is usually ready within moments of first launch; the writing model and Enhanced Text-to-Speech become available once their downloads finish.

After everything is downloaded, you can disconnect from the network entirely. All generation is local and fully offline from that point on.

Where models are stored

Downloaded models live in Arqenne's application support folder:

~/Library/Application Support/com.arqenne.studio/

Inside that folder:

Location	Contents
`tts/`	Standard Text-to-Speech engine and its voices
`models/`	Enhanced Text-to-Speech and the AI writing model

You do not need to open these folders for normal use — Arqenne manages them for you. They are listed here so you can find them if you ever want to check disk usage or reclaim space (see Reclaiming disk space below).

How much disk to expect: plan for roughly 4 GB total once all three model sets are downloaded. The bulk of that is Enhanced Text-to-Speech (~3.5 GB). If you only use Standard Text-to-Speech and the writing model, expect closer to 600 MB.

System requirements

An Apple Silicon Mac. Arqenne is built for Apple Silicon (M-series) Macs.
Standard Text-to-Speech is light enough to run on any supported Mac.
Enhanced Text-to-Speech and the AI writing model use the GPU (Metal) for fast, on-device generation.
Memory affects the writing model: Macs with 12 GB or more get a larger, higher-context variant automatically.

Generated audio is 24 kHz WAV in every case.

Managing models

If a download is interrupted

If you quit Arqenne, lose your connection, or restart mid-download, nothing is lost. The next time the app launches (or the next time you use the feature), Arqenne picks the download back up automatically. Files that already finished are kept; only what is missing is fetched.

If a download seems stuck rather than simply slow, quit and relaunch Arqenne — see Troubleshooting below.

Reclaiming disk space

If you are tight on disk and want to recover space, you can remove a model by deleting the relevant folder inside Arqenne's application support directory:

To remove…	Delete this folder
Standard Text-to-Speech	`~/Library/Application Support/com.arqenne.studio/tts/`
Enhanced Text-to-Speech and the writing model	`~/Library/Application Support/com.arqenne.studio/models/`

Deleting a model is reversible, but it is not free. The next time you use a feature that needs that model, Arqenne re-downloads it automatically — which means a network connection and another download of the same size. Only remove a model if you genuinely need the space and do not mind fetching it again later.

Removing the models/ folder takes back the most space (Enhanced Text-to-Speech is ~3.5 GB), but it is also the slowest to re-download. Removing tts/ only frees ~90 MB.

The models/ folder holds both Enhanced Text-to-Speech and the AI writing model. Deleting it removes both; each is re-downloaded the next time you use the feature that needs it.

Troubleshooting

A download is stuck or failed

Check your network connection. The first download of each model needs internet access.
Quit Arqenne and relaunch it. Downloads resume from where they left off — you will not re-fetch what already completed.
If it still will not progress, confirm you have enough free disk space (see below), then try again.

Not enough free disk space

Make sure you have headroom for the model you are downloading — about 4 GB free if you want all three sets. Free up space (the Storage section of System Settings is a good place to start), then relaunch Arqenne so it can finish the download.

A feature is "still loading" right after launch

When Arqenne starts, it loads its models into memory before they are ready to generate. For a moment after launch, a feature — narration or AI writing suggestions — may show as loading or be temporarily unavailable. This is normal and usually clears within a few seconds. Wait a moment and try again. If a model is still downloading (first launch only), the feature becomes available once that download finishes.

A feature never becomes ready

If a feature stays unavailable well past launch, quit and relaunch Arqenne. If the problem persists, deleting that model's folder (see Reclaiming disk space) forces a clean re-download on the next launch.

Privacy

Models download once, from a model host, the first time you launch Arqenne or first use a feature.
After that, nothing is sent anywhere. All narration, voice cloning, and writing suggestions are generated on your Mac. Your manuscripts, your text, and your audio never leave the device.
You are in control of the files. The models live in a folder you can inspect or delete at any time; Arqenne re-downloads only what you ask it to use again.