I’m Eduarth Schmidt—Product Designer, UX/UI Designer & AI Innovator. Over the last months I stress-tested the newest AI image capabilities inside Figma against a traditional round‑trip into a full image editor. My goal: figure out when I can stay on the canvas and ship, and when I still need pixel‑level control.
Short answer: I stay in Figma’s AI image tools for most UI tasks (hero sections, cards, onboarding visuals, marketing blocks), and I switch only when brand‑critical precision or heavy retouching is required. Below I’ll show you the exact scenarios, the time savings, and the pitfalls to watch for.
The Core UX Problem: Context Switching
Every detour to a separate editor costs momentum: export → edit → re-import → resize → re-check. That loop fragments attention and delays iteration—especially in fast‑moving product teams. In-canvas AI editing removes the tool‑hopping so I can keep layout, typography, and image decisions tightly coupled.
Why this matters: When the layout is alive and the team is in the file, edits happen in minutes instead of micro‑handoffs that take hours.
What Figma’s AI Image Tools Actually Do
Figma’s new image editing focuses on three actions I use constantly in UI work. Think of them as layout‑first image operations.
Erase (Object/Background)
- Best for: Removing small distractions, cleaning up stock images for cards/banners, or clearing busy backdrops behind a product.
- What I watch for: Hair, glass, and fine edges can produce artifacts. For app UI, it’s often “good enough,” but for brand photography I verify at 100% zoom.
Isolate (Auto Subject Cutout)
- Best for: Quick product/person cutouts for avatars, feature callouts, and modular marketing blocks.
- What I watch for: Limited manual refinement. Great for speed; not ideal for surgical masks.
Expand (Outpainting)
- Best for: Adapting images to new aspect ratios, creating safe space for headlines/buttons, and avoiding awkward crops across breakpoints.
- What I watch for: It respects composition well for simple scenes. For complex textures, I sanity‑check seams.
Speed Test on Real UI Tasks
I timed typical UI scenarios using in‑canvas AI editing versus a full editor round‑trip. Times are averages from repeated runs on production‑like assets.
Task 1: Landing Page Hero
- In‑canvas AI: ~2–3 minutes to isolate subject, expand background for headline space, and erase clutter.
- Round‑trip: ~10–15 minutes including exports, masks, cleanup, and re‑import.
- Verdict: Stay in‑canvas. You get 80–90% quality at ~20% of the time, which is perfect for iterative product UI.
Task 2: Product Cutout With Crisp Edges
- In‑canvas AI: Fast cutout, occasional halos on tricky edges (hair, soft shadows).
- Round‑trip: Precise masking, path control, edge refinement.
- Verdict: Switch when the edges sell the story. For ecommerce or brand‑grade imagery, I still go for pixel‑level control.
Task 3: Multiple Aspect Ratios (Responsive/Ads)
- In‑canvas AI (Expand): One‑click extensions with usable results for 1:1, 16:9, 4:5.
- Round‑trip: Manual content‑aware operations and checks.
- Verdict: Stay in‑canvas. It’s built for layout adaptability.
Task 4: Heavy Retouching & Art Direction
- In‑canvas AI: Not designed for advanced compositing or meticulous color work.
- Round‑trip: Full control of layers, masks, and tone.
- Verdict: Switch. When the image itself is the product, I want the deep toolkit.
The 80/20 Rule for UI Teams
In day‑to‑day product design, about 70–80% of my image edits are layout‑driven: make space for text, clean a background, cut a subject, adapt a ratio. Those are now comfortably handled in‑canvas. The remaining 20–30% are brand moments or campaign assets where quality trumps speed—those still warrant a specialist edit.
Practical takeaway: Reframe the question from “Which tool is better?” to “Which edits are layout‑first vs. image‑first?” If the layout leads, stay. If the image leads, switch.
My “No‑Round‑Trip” UI Workflow
This is the pattern I’m using with product teams to ship faster while keeping visuals on‑brand.
- Drop the raw image on the canvas. No pre‑prep.
- Isolate the subject into its own layer.
- Expand the scene to your target aspect ratio(s). Create paddings for headline, CTA, and safe areas.
- Erase minor distractions (stray objects, seams, tiny logos) so typography reads clean.
- Audit at 100% on a neutral background. If artifacts are visible in key areas, branch the workflow for a specialist pass.
When to branch:
- Hair, glass, or soft‑focus edges on hero imagery.
- Brand‑critical campaigns where color, detail, and texture fidelity matter.
- Complex composites with shadows and reflections.
Quality Checks I Never Skip
- Edge Hygiene: Zoom to 100%. Look for halos or compression bursts around the subject.
- Text Safety: After Expand, verify readable space for headings/buttons across breakpoints.
- Consistency: Keep visual style consistent across cards/sections—don’t let AI introduce mismatched lighting.
- Accessibility: Ensure contrast ratios still pass when backgrounds shift. I validate key components against WCAG contrast guidelines.
Team Play: Why In‑Canvas Edits Improve Collaboration
Editing on the same canvas lets PMs, marketers, and engineers comment on the actual layout. Instead of debating screenshots, we iterate on real components with real spacing. The result? Fewer files, fewer exports, fewer mismatches at handoff.
Tip: Use named versions to capture pivotal visual directions without branching into file chaos.
Decision Matrix: Stay or Switch?
Stay in‑canvas when:
- You’re shipping UI fast and need usable visuals now.
- The goal is layout adaptability (aspect ratios, safe areas, responsive fit).
- Artifacts are imperceptible at typical UI display sizes.
Switch when:
- Precision masking defines quality (hair, fur, translucent materials).
- Heavy retouching, color grading, or complex compositing is part of the ask.
- You’re producing campaign‑grade or print‑grade assets.
Final Take
“As a Product Designer and UX Specialist, my job isn’t to worship tools—it’s to protect momentum. If an edit serves the layout, I stay in Figma. If the image carries the brand story, I switch without hesitation.”
— Eduarth Schmidt, Product Designer & UX Specialist
In 2026, in‑canvas AI image editing replaces friction, not craft. For UI‑first work—where composition, copy, and component behavior move together—I stay on the canvas and ship confidently. When the image itself carries the brand moment, I still reach for deeper control.
Design faster. Switch tools less. Keep momentum.
Explore more on AIAutomationFlows.com.
