Apple unveiled a full suite of photorealistic AI photo editing tools at WWDC 2026 on June 9, reversing the cautious stance the company held when it first introduced Apple Intelligence in 2024. Image Playground can now generate photorealistic images using a real person as a reference. Three new Photos app features — an upgraded Clean Up, plus Extend and Spatial Reframing — let users rewrite their pictures by tapping, brushing, or describing what should change. Every edit gets stamped with a SynthID watermark.
The pivot is sharp. Apple shipped Clean Up two years ago as a Magic Eraser equivalent that quietly scrubbed objects from photos, and stopped there. Google Photos and Samsung's Galaxy AI moved well past that line, letting users add explosions, props, and entire scene elements via prompt. Apple sat out, and Craig Federighi gave the reason on the record.
When Apple Intelligence was announced in 2024, Federighi said the company was concerned AI could change how people view photographic content as something they can rely on as indicative of reality. A separate line from the Clean Up rollout was blunter.
Key facts
- 01Apple unveiled photorealistic AI editing across Image Playground, Clean Up, Extend, and Spatial Reframing at WWDC 2026 on June 9.
- 02Every photo edited with Apple Intelligence will carry a SynthID watermark, Google's near-invisible AI-content tag.
- 03The shift reverses a 2024 stance from Craig Federighi, who warned AI could change how people view photographs as 'indicative of reality.'
- 04Clean Up, launched two years ago as a Magic Eraser competitor, gets a 'major upgrade' with more realistic infill on complex scenes.
- 05Spatial Reframing reuses perspective models Apple built for Vision Pro to let users drag photos into new camera angles.
That position is now retired. The updated Image Playground generates images in a photorealistic style for the first time and accepts natural-language prompts or direct gestures — tap, circle, or brush — to move and resize objects. In Apple's keynote demo, a real photo of a woman was used as input, with a prompt asking the model to put a chocolate cake in her hands. The system added the cake and also replaced her outfit and the background without being asked.
Apple framed the expansion in characteristically polished language.
“offers new powerful ways for users to bring their imagination to life.”— Apple, company statement
Clean Up has been given what Apple calls a major upgrade, with better quality and more realistic infill on complex scenes. Extend mirrors Adobe Photoshop's Generative Expand, filling blank canvas beyond an image's original borders so a portrait can be reshaped into a landscape. Spatial Reframing borrows from the depth and scene-understanding work Apple built for Vision Pro, letting users drag part of a photo to adjust the apparent camera angle and generating new pixels only where the perspective shifts.
The guardrail is SynthID, the watermarking system Google introduced for its own generative tools. Apple says any photo touched by Clean Up, Extend, Spatial Reframing, or Image Playground will be embedded with SynthID so downstream platforms can flag it as AI-manipulated. SynthID adoption widened earlier this year when OpenAI signed on, and users can verify a file by uploading it to Gemini or Google Search and asking whether the watermark is present. Apple had previously relied on a proprietary metadata tag for Clean Up edits that, by the company's own admission, no other major platform read.
The trade is straightforward: Apple gets to compete with Google and Samsung on prompt-driven editing, and SynthID carries the weight of distinguishing real from synthetic. Whether that holds depends on platform-level enforcement. Automatic SynthID labeling across social networks is in early stages, and the bulk of deepfaked and AI-generated imagery online still circulates without any watermark at all.
The product logic is hard to argue with. Apple's hardware and on-device silicon are well suited to running these models locally, and iPhone owners are already using competitor tools to do the same edits with worse provenance signals. Shipping Extend and Spatial Reframing inside Photos, with watermarking turned on by default, is arguably a cleaner outcome than ceding the workflow to apps that disclose nothing. The Vision Pro spatial work also finally has a consumer use case that ships at scale, rather than living inside a $3,499 headset.
The harder question is the one Apple itself raised in 2024 and hasn't re-answered. If a photo's background, outfit, and camera angle can all be rewritten in seconds, the file is still called a photo inside the Photos app, and the SynthID tag is invisible to a human eye, the burden of distinguishing memory from generation lands on downstream platforms that mostly haven't built the infrastructure yet. Apple is betting that watermark adoption catches up before public trust in personal images runs out.
For the AI market, this closes one of the last remaining holdouts. Every major consumer platform — Google, Samsung, Apple, plus OpenAI's image tools — is now shipping prompt-driven photo manipulation, and SynthID is becoming the de facto industry standard for provenance. That consolidates Google's position on watermark infrastructure in a way that's easy to miss in the WWDC noise: Apple just plugged its photo pipeline into a Google-built system, which gives DeepMind real leverage over how synthetic-media labeling evolves across the iPhone install base.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
