Affinity Generative Fill: Free App, Paid AI

28:41

Affinity by Canva is genuinely free — the whole professional application, on Mac and Windows, with no limits on the design tools and no payment. That part is not a trial or a teaser.

The generative AI is the part that is not free, and the confusion is understandable: it appears inside the same application, in the same menus, alongside tools that cost nothing. So the practical question for anyone who downloaded Affinity expecting a free Generative Fill is what the AI actually costs, and what the alternatives are.

This guide answers both, and is specific about where each option is genuinely better — including where the built-in tools win.

What do you get free in Affinity?

The complete application. Canva describes the new Affinity as free forever with the full professional experience — vector, photo and layout tools in one app, non-destructive editing, RAW development, retouching, artboards, CMYK print output and full export, with no feature gate and no payment.

It also opens the formats you are likely to have lying around: PSD, AI, PDF, SVG, TIFF and IDML among others. For anyone leaving a subscription editor, that alone removes most of the migration friction.

What is not included is the generative layer, which is worth stating plainly because it is the one thing people arrive looking for.

Does Affinity have Generative Fill?

Yes, but behind a subscription. Generative Fill, Expand & Edit and Remove Background are provided by Canva AI Studio inside Affinity, and Canva's own announcement states they are available to everyone with a Canva premium account. The free tier of Affinity does not include them.

This is a meaningful improvement on where Affinity was before — it previously had no generative fill at all, and a lot of older articles still say so. If you have seen a guide claiming Affinity cannot do generative fill, it is out of date.

Two things worth knowing about the built-in tools:

• They work inside your document. Make a selection, run Generative Fill, and the result lands on the canvas in context. Nothing leaves the app and nothing has to be re-imported. That is a real advantage and it is the main reason to pay.
• Canva states your Affinity work is not used to train AI features. That is Canva's own commitment in the launch announcement, and it is a reasonable question to ask of any AI tool you point at client work.

Pricing for Canva's premium plans varies by region, plan and seat count and has changed over time, so check Canva's current pricing page rather than any figure printed in an article — including this one, which deliberately does not quote a number.

Can you install AI plugins in Affinity?

You can install Photoshop-style filter plugins — the .8bf kind, added through Preferences — and popular AI filters such as Topaz, DxO and the Nik Collection work that way. What does not exist is a third-party generative panel where you type a prompt. The prompt-driven tools inside Affinity come from Canva.

The distinction matters more than it first appears, and it is where a lot of advice goes wrong.

• Filters are supported. A .8bf plugin receives pixels and returns pixels. Plenty of those are AI-powered — denoise, sharpening, upscaling — and they run inside Affinity without any subscription beyond the plugin's own licence.
• Prompt panels are not. The kind of plugin where you type a description and a model generates something back has not appeared for Affinity the way it has for Photoshop. Generative tooling inside Affinity is Canva's.

So "Affinity has no AI plugins" is wrong, and "you can just install an AI generator plugin" is also wrong. The accurate version is that you can extend Affinity with filters, and generation happens either through Canva or outside the application entirely.

What are your options for AI with Affinity?

Three, and they are genuinely different rather than three versions of the same thing: pay for Canva premium and work inside the document, run models outside Affinity and move pixels on the clipboard, or use free web tools and accept watermarks, caps and uploading your work to a browser.

Feature availability changes; Canva's tool list is from its Affinity launch announcement, checked August 2026.

What does each route actually cost?

A subscription is a fixed monthly amount regardless of how much you generate. Running models yourself is priced per image — roughly three to fifteen cents depending on the model and resolution — with nothing charged in months you do not generate anything. Which is cheaper depends entirely on your volume.

Rather than pretend one wins, the honest way to decide is to work out where they cross. Take whatever the subscription costs you per month and divide it by your per-image rate. That is roughly how many images a month you need to generate before the subscription is the better deal.

Rates from Rangy's live pricing tables at the cheapest configured provider, checked August 2026. Providers change their rates, and 1K is cheaper while 4K costs more.

The two shapes suit different people. Steady daily generation favours a flat fee. Spiky, project-driven work — a heavy week, then nothing for a month — favours paying per image, which is the same reasoning set out in the guide to using AI image generation without a subscription.

How do you run AI models outside Affinity?

Through a desktop app that holds your own provider API keys and generates locally-saved files. You pick the model, it charges the provider's rate per generation, and the output lands on disk as a normal image ready to bring into Affinity. Nothing is uploaded to an intermediate service.

What that setup buys you over the built-in tools is choice and cost visibility. You are not limited to one provider's curated model list, you can run two models on the same prompt and compare, and you can see what each generation costs before committing to it.

The pieces of the workflow that Affinity users tend to actually use:

• Clipboard in, clipboard out. The round trip below. This is the part that makes an external app feel less external.
• Upscaling. Crystal handles 2x to 10x, which covers taking a generated element up to print size before it goes into an Affinity layout.
• Transparent output. GPT Image 1.5 can return a native alpha PNG directly, so a generated element arrives already cut out rather than needing masking in Affinity.
• Skin and detail work. A dedicated skin enhancer rebuilds texture when a generated portrait comes back waxy, which is the failure mode covered in the portrait retouching guide.

How do you get the result back into Affinity?

On the clipboard, in both directions. Copy a layer in Affinity, paste it into the AI app as the input image, generate, then send the result back — it copies to the clipboard, brings Affinity to the front and sends the paste keystroke, so the image arrives as a new layer without a file export in either direction.

A few practical notes, because this is the part where expectations and reality can diverge:

• It arrives as a layer, not inside a selection. The automated step sends an ordinary paste. If you want the pixels constrained to an active selection, use Affinity's own Paste Inside afterwards — Cmd+Option+V on macOS, Ctrl+Alt+V on Windows.
• macOS needs Accessibility permission for the app to send a keystroke to another application. Without it the image is still on your clipboard and you paste manually; the workflow degrades rather than breaks.
• It works with the current free Affinity. The integration looks for Affinity.app first — the Canva-era single application — and falls back to the older Affinity Photo installs.
• You are moving flattened pixels. The clipboard carries a rasterised image. Layers, masks and adjustments stay in Affinity, which is where they should be edited anyway.

When is the built-in AI the better answer?

When the work is inpainting inside a document. Masking a selection, filling it in context and continuing to edit is genuinely smoother in one application than across two, and no amount of clipboard convenience closes that gap. Volume matters too — heavy daily use makes a flat fee cheaper.

The clear cases for paying for Canva premium instead, stated directly because this guide is published by a company that sells the alternative:

• Frequent masked fills. Removing a lamppost, extending a background edge, patching a blemish region. Selection-based work belongs where the selection lives.
• High volume. If you generate every day, per-image pricing stops being an advantage and the flat fee wins on both cost and friction.
• You already pay for Canva. If a premium plan is already in your stack for other reasons, the AI is included and there is little reason to add another tool.
• You want one application. A perfectly reasonable preference. Every external route adds a window, a set of API keys and a step.

And one case for neither: if all you need is an occasional background removal, Affinity's selection tools and a .8bf filter plugin will handle a lot of it without any generative model at all.

Frequently asked questions

The bottom line

Affinity is free and the AI inside it is not, so the decision is about cost shape rather than capability. Heavy, selection-based generative work justifies a Canva premium plan. Occasional or project-driven generation is cheaper per image outside the app, moved across on the clipboard.

What has changed since most articles on this subject were written is that Affinity is no longer the app without any generative AI. It has Generative Fill, Expand & Edit and Remove Background — they simply sit behind a plan, the way most generative tooling now does.

So the question is not whether Affinity can do it. It is whether a recurring plan or a per-image charge fits how often you actually generate.