AI editors and model exports
| Editor or model family | What you can use | What stays outside the cleanup |
|---|---|---|
| ChatGPT / OpenAI images | Saved JPEG, PNG or WebP. Supported attached Content Credentials containers can be removed. | Invisible watermarks. No certification of every OpenAI model or export path. |
| Gemini / Google images | JPEG, PNG or WebP image exports. Supported hidden file information is cleaned. | SynthID in the image itself, visible corner marks, and AI detection. |
| Photoshop / Firefly | Final JPEG, PNG or WebP exports. Supported EXIF, XMP, IPTC and C2PA containers. | PSD, PSB, layered projects, visible marks and durable watermark signals. |
| Midjourney, FLUX, Stable Diffusion & other editors | The same supported image formats. No model-specific connection is needed. | Export contents vary by app and workflow. No source-by-source or version-by-version success claim. |
The names above identify possible source workflows, not partnerships or individually verified label-removal results. A valid image can also contain no removable information.
Primary references: OpenAI provenance, Google SynthID, Adobe Content Credentials.
Supported file types
Supported attached application metadata and comments are removed. Compressed image data and color profiles stay.
Supported text, EXIF and provenance chunks are removed. Pixels, transparency and supported animation stay.
Supported EXIF, XMP and other removable chunks are cleaned. Image, alpha, color and supported animation stay.
Supported metadata fields are cleared in standard containers. Video, audio and playback offsets stay unchanged.
Export first: HEIC, AVIF, RAW, GIF, TIFF, PSD, SVG, audio and text are not direct cleanup inputs. Photo files are limited to 25 MB each, with up to 20 per batch. Touch devices use a 60 MB batch budget; other devices use 150 MB. Video is one clip at a time, up to 50 MB on touch devices or 100 MB otherwise.
MP4/MOV support has structural limits: fragmented, encrypted and unsupported extra-track files are rejected. A recognized extension alone does not establish support. See the video tool and its limits.
Where are you posting?
Try a cleaned copy before a new post.
PHOTO CLEANUP
TikTok
What file cleanup can change, and why existing auto-labels stay.
AUTOMATIC LABELS
YouTube
Clean a clip. Check AI disclosure separately in Studio.
VIDEO & DISCLOSURE
The photo workflow for Facebook uploads.
META PHOTO GUIDANCE
Threads
The photo workflow for your Threads post.
META PHOTO GUIDANCETransfa prepares a local file; it has no posting integration or access to your social accounts. One early user reported an Instagram photo result. Facebook, Threads, TikTok and YouTube label-removal outcomes have not been validated. Existing posts are never changed.
How your files are processed
The cleaner is TypeScript running in a Web Worker in your browser. It parses supported file containers instead of generating a replacement image. There is no model API call, AI credit balance or background agent editing your file.
Our VPS serves the website and an anonymous counting service. Google Analytics and PostHog measure site actions as described in Privacy & analytics; they do not receive your media files. The only time a cleaned file leaves the page is when you choose to save or share it through your browser or another app.