Calabi Labs · Guide · 2026-06-13
How to clean up photos on an iphone remove people and objects from photos
```html
Your iPhone Can Now Visually Remove People and Objects From Photos — Here's How
Starting with iOS 18.1, Apple built a feature called Clean Up directly into the Photos app that lets you circle or brush over people and objects to remove them from your photos using on-device AI. If you're looking for a quick, free way to visually erase unwanted elements from a photo taken on your iPhone, that's the tool to use.
But if you're working with AI-generated images — photos created with tools like Midjourney, DALL-E, Flux, Stable Diffusion, or Sora — and you want to post them without platforms flagging them as AI content, that's a different problem entirely. Your iPhone's Clean Up tool won't help with that. That's where Calabi comes in.
What Actually Gets Your Photo Flagged as AI-Generated
When you export an image from an AI generator, it carries an invisible metadata layer that platforms like Instagram, TikTok, YouTube, and Reddit scan automatically. This isn't about how the image looks — it's about the file's hidden structure. Several specific signals trigger detection:
- C2PA / Content Credentials: AI tools like Midjourney and OpenAI embed JUMBF atoms — cryptographic manifests — directly into the file that declare "this was made by AI." These survive cropping and re-encoding in many cases.
- DigitalSourceType: trainedAlgorithmicMedia: An XMP metadata tag that explicitly flags an image as produced by a trained AI model. This is a direct red flag for platform scanners.
- Generator and encoder fingerprints: AI export pipelines use specific encoders — think Lavc, x264, FFmpeg — that leave traceable SEI (Supplemental Enhancement Information) markers in video files, or similar structural fingerprints in images.
- Missing capture metadata: A real phone photo has GPS coordinates, a capture timestamp, a device Make/Model, and software version. AI exports have none of this. That absence itself is a signal.
Platform scanners in 2026 check all of these simultaneously — C2PA manifests, XMP AI flags, encoder fingerprints, and the absence of authentic phone-capture metadata. A visually clean image can still fail these checks in seconds.
Why the Obvious Fixes Don't Work on AI Export Metadata
If you've tried uploading an AI-generated image and gotten flagged, you've probably already attempted some workarounds. Here's why each one falls short:
- Screenshotting the AI image: This removes some metadata but adds a new problem — your screenshot carries your own device's capture metadata, which still won't match a real camera and often still contains traces of the original file.
- Cropping heavily: C2PA manifests and XMP tags survive most crops because they're embedded at the file level, not tied to specific pixels. You can crop away the visible watermark and still trigger detection.
- Re-exporting or saving to a new format: Many platforms and tools re-compress on save, but the AI metadata layer often persists because the structural tags weren't actually removed — just visually hidden.
- Using iPhone Clean Up on an AI image: Apple's Clean Up tool performs visual inpainting — it reconstructs the pixel content where you mark a region. This changes how the image looks but does not strip the AI metadata layer. The file will still carry its C2PA and XMP AI flags after editing.
The detection layer and the visual layer are independent problems. You need a tool that addresses the file-level metadata, not just what's visible in the frame.
How to Actually Clean an AI Image's Detection Metadata
Calabi is a one-pass web tool that strips the invisible detection signals and injects authentic phone-capture identity, so your AI-generated image passes as a normal phone photo at the file level. Here's the process:
- Upload your AI-generated image — any format the platform accepts. No manual settings, no selecting regions.
- Calabi's pipeline runs automatically: it strips every C2PA/JUMBF manifest, removes the DigitalSourceType: trainedAlgorithmicMedia XMP tag, clears generator and encoder fingerprints, and eliminates all other AI-signature metadata.
- Calabi injects authentic phone identity: it writes real device profiles — iPhone 15 Pro, Pixel 8 Pro, Galaxy S24 Ultra — including Make, Model, software version, GPS coordinates, and a genuine capture timestamp, along with a real-phone encoder name.
- You review the forensic proof card before downloading: this is an ExifTool-readable report showing exactly what was stripped and what was injected, so you can see precisely what changed.
- Download the cleaned file and post it directly to your platform of choice.
The result is a file that carries all the metadata signals of a real phone photo, with zero C2PA or AI-flag tags — without altering a single pixel of the image itself.
Frequently Asked Questions
Can I use iPhone Clean Up and Calabi on the same image?
Yes. If you've used Apple's Clean Up tool to visually remove an unwanted person or object from an AI-generated image, you can then run that edited image through Calabi to strip the AI metadata layer. Calabi handles the file-level signals; Clean Up handles the visual content. They solve different problems.
Does Calabi remove visible watermarks like the Sora sparkle or Midjourney corner logo?
Calabi does not edit pixels, so it won't erase a visible logo from the frame. If you need to remove a visible watermark, crop it out using any photo editor first — then run the cropped image through Calabi to strip the AI metadata that would still get you flagged even after cropping.
Which platforms scan for AI metadata?
Instagram, TikTok, YouTube, and Reddit all run automated scans on uploads — often within seconds of posting. They check C2PA manifests, XMP AI flags, encoder fingerprints, and the presence or absence of phone-capture metadata. A cleaned file will present as a normal phone photo to these scanners, but results vary by platform and source model.
Try Calabi free at calabilabs.com — 10 cleans, no card. ```