Calabi Labs · Guide · 2026-06-17
1remove people from photos app
The search query "1remove people from photos app" tells me exactly what you want: an app that erases a person from a photo. Here's the direct answer.
There are solid inpainting apps — Cleanup.pictures, Inpaint, Adobe Photoshop's Remove Tool, and Apple Intelligence's Clean Up in the Photos app — that let you brush over a person and the AI fills in the background. They work pixel-by-pixel, reconstructing what was behind that person. If your goal is purely visual, those tools do the job.
But if you're asking because you want to post the result on Instagram, TikTok, or YouTube without getting flagged or suppressed — that's a different problem entirely. And it's not solved by erasing pixels.
What actually gets your file flagged — and it's not the pixels
When you upload a photo, platforms don't just look at what it looks like. They scan the invisible metadata layer baked into the file itself. Here's what's actually triggering detection in 2026:
C2PA / Content Credentials — This is a cryptographic manifest embedded in the image file that declares "I was made by AI." It lives in JUMBF boxes inside JPEG and PNG files, and it's becoming standard in exports from Midjourney, DALL-E, Sora, Runway, and most AI image generators. A single AI export can contain 18 or more of these JUMBF atoms. Platforms like Reddit and YouTube already scan for them.
XMP AI flags — Specifically the DigitalSourceType: trainedAlgorithmicMedia tag. This is an XMP metadata field that explicitly identifies a photo as AI-generated. It's separate from the visual content entirely.
Encoder fingerprints — When you export from an AI tool, the video codec or image encoder adds specific fingerprints to the file. Lavc (FFmpeg's encoder) and x264 SEI messages in video files are dead giveaways. These signals survive cropping and screenshotting.
Missing GPS, inconsistent timestamps, and device profiles — A real phone photo has GPS coordinates, a capture timestamp in a specific format, a real Make/Model (iPhone 15 Pro, Pixel 8 Pro, Galaxy S24 Ultra), and a genuine encoder name. AI exports have none of that — or they have it wrong. Platforms flag the absence of these signals as suspicious.
Even if you use an inpainting app to visually remove a person from your AI-generated photo, every one of those invisible signals is still in the file. The person is gone. The "made by AI" fingerprint isn't.
Why cropping, screenshotting, and inpainting don't fix it
Here's the practical problem: apps like Cleanup.pictures, Inpaint, and Photoshop work on the visual layer — the pixels you see. They do a genuinely good job of reconstructing the background where a person was. But they don't touch the metadata layer underneath.
Screenshotting a photo strips some metadata, but it also introduces new compression artifacts and still leaves XMP AI flags and C2PA manifests intact in the original file you're screenshotting. Cropping removes visible content, but the JUMBF boxes and encoder fingerprints are embedded in the entire file structure — cropping the visible image doesn't remove them.
Inpainting tools have no mechanism to strip C2PA manifests, remove DigitalSourceType: trainedAlgorithmicMedia from XMP, or inject a real device profile with GPS coordinates and a genuine capture timestamp. You're editing pixels. The file's digital身份证 is still AI.
How Calabi actually handles it — the metadata layer
Calabi is a one-pass web tool that works on the invisible layer. You upload your AI-generated image — after you've used whatever inpainting tool you prefer to clean up the visuals — and Calabi runs a pipeline with three stages:
Strip — Remove every detectable AI signal: JUMBF / C2PA atoms (reduced from 18+ to 0), the DigitalSourceType: trainedAlgorithmicMedia XMP flag, generator/tool tags, and encoder fingerprints like Lavc and x264 SEI. A raw AI export carries roughly 144 metadata tags; Calabi reduces that to about 94 neutral structural tags.
Inject — Write in a real phone identity: iPhone 15 Pro, Pixel 8 Pro, or Galaxy S24 Ultra — with real Make, Model, Software version, GPS coordinates, and a capture timestamp. This makes the file look like it came from an actual device, not an AI generator.
Verify — Return a forensic proof card showing exactly what was stripped and what was injected. This is the same ExifTool scan that platforms and newsrooms use, so you can see what they'll see before you post.
The visual edit — the inpainting — is still yours to do in your preferred app. Calabi handles the file-level identity so the result doesn't get flagged at upload.
Step-by-step: how to use both together
- **Visually clean the photo first** — Use Cleanup.pictures, Inpaint, Photoshop, or Apple Clean Up to brush over the person you want removed. Get the visual result you want.
- **Upload to Calabi** — Drop the cleaned file. The pipeline runs automatically — no settings to adjust, no tool to select.
- **Review the forensic proof card** — You'll see exactly which AI signals were stripped and which phone identity was injected. ExifTool shows you what the platform will see.
- **Download and post** — The cleaned file is ready for Instagram, TikTok, YouTube, or Reddit without the metadata layer that triggers automatic scans.
FAQ
Can Calabi remove a person from my photo like Cleanup.pictures does? No. Calabi doesn't edit pixels, reconstruct backgrounds, or use inpainting. It works on the metadata and encoding layer. Use an inpainting app for the visual edit, then Calabi to clean the file's digital identity afterward.
Will platforms still detect my AI image after using Calabi? Calabi removes the documented metadata signals platforms scan for — C2PA manifests, XMP AI flags, encoder fingerprints. Results vary by platform and by how the original image was generated, but these are the specific signals Calabi verifies as fully removed with ExifTool.
Do I need to crop or screenshot my AI image before using Calabi? No. Calabi strips embedded metadata regardless of visible content. Cropping doesn't remove JUMBF boxes or encoder fingerprints — and it may reduce image quality. Use your inpainting tool on the full image, then run it through Calabi.
What if my AI image has a visible watermark in the corner? Calabi doesn't erase logos or watermarks from the visible image. If a visible Sora sparkle or Midjourney badge is present, cropping it out or using an inpainting tool to remove it is the right call. Note that the invisible metadata layer — C2PA manifests, XMP flags, encoder fingerprints — survives cropping, which is exactly what Calabi removes.
Try Calabi free at calabilabs.com — 10 cleans, no card.