AI Features

Blitter includes two optional features that use an AI service to turn a plain-English description into something usable in the app. Both are off until you set up a provider and API key in Settings → AI Providers, and neither runs on its own — each only makes a request when you type a description and press Generate.

• Director — describe a mood, use, or feeling (“CRT terminal boot”, “rain on a window at night”, “anxious / jittery”), and it proposes a whole animation setup: Sequence, Time shift, colors, and timing.
• Describe a shape — describe a shape in words, and it draws a custom grid shape you can use as a mask, the same as one you’d draw by hand.

Bring your own key

Blitter has no server and no account of its own. You choose which AI service to use, enter an API key for that service, and requests go directly from your Mac to the provider you selected, authenticated with your key and billed to your account.

One provider is active at a time, and it applies to both features.

How keys are stored

Your API keys are stored in this Mac’s Keychain, readable only by your user account. Blitter keeps one key per provider, so switching between them doesn’t mean pasting again, and only the provider behind the selected model is ever contacted. Blitter never receives your prompts. No key, model id, or endpoint is written into .blit files or your Gallery — including anything you share to the Community Gallery.

See the Privacy Policy for the full picture of what’s sent and when.

Providers and models

Three providers are available. Each is curated with a few suggested models and a live cost estimate; you can also type any model id that provider supports, so a newly released model is a Settings change, not an app update.

Each provider you add gets its own row in Settings with a “Get an API key” link, so you don’t need to hunt for the right page.

Anthropic keys and workspaces

Anthropic keys come in two kinds, and one of them needs a second piece of information. A key scoped to a single workspace works on its own — paste it and you’re done. A key linked to your identity that can act in more than one workspace has to say which workspace each request belongs to, and Anthropic rejects the request otherwise with a message naming anthropic-workspace-id.

If you have one of those, paste the workspace’s ID into Workspace ID in Settings. It starts with wrkspc_ and is listed in the Claude Console under Settings → Workspaces. Leave the field empty for a single-workspace key: sending an ID with a key that is already scoped to one workspace is itself an error.

What it costs

Each generation is one request — a short prompt and instructions in, a small JSON answer back. At the curated models’ published rates that’s typically a fraction of a cent to a few cents, depending on the model. Settings shows the input and output rates and an estimate for the model you’ve selected. A model id you type yourself shows a link to that provider’s pricing instead, since Blitter doesn’t know the rate.

Blitter keeps a count of the requests it has sent from this Mac. Settings shows that count for the current month and for all time. The count stays on this Mac. The provider charges the real amount, and published rates change.

The monthly warning is off until you turn it on under Usage. It starts at 200 requests. Once this month’s requests reach that number, the next Generate or Test stops and asks. You can send that one request, raise the limit, or cancel. Next to the number, Blitter shows what that many requests would cost on the model you have selected.

The limit that matches the bill is the one you set with the provider. Usage links to each provider’s spending-cap page.

Setting it up

1. Open Settings and find AI Providers.
2. Click Add provider and choose one. A row appears for it.
3. Follow that row’s “Get an API key” link, create a key on the provider’s site, and paste it into the row’s API key field. Once saved it shows as dots; press Edit to replace it, then Save.
4. Click Test in the row. It makes one tiny request and reports either “Working” or exactly what the provider objected to — the quickest way to find out that an account has no credit, or that a key needs a workspace ID, before you’re mid-prompt.
5. In the Model section above, choose the model you want to use. Picking a model also picks its provider, and the rates and estimated cost per generation are shown right beneath it.
6. Open Director or describe-a-shape and try a prompt.

You can add as many providers as you like and switch between them by changing the Model selection; each keeps its own key. Remove Provider deletes that provider’s key from the Keychain and returns it to the Add provider menu, so you’ll need to paste a key again to use it later.

See the Privacy Policy for what’s sent to the provider, how keys are stored, and each provider’s own terms and privacy policy.