1. The first run is slow — by design, not by accident
Browser-based background removal tools load a neural network model into your browser before doing any processing. Depending on the engine you choose, that model is between 5 MB and 176 MB. It needs to download once.
On a fast connection, the General engine (~80 MB) loads in 10—20 seconds. On mobile data or a slow Wi-Fi connection, it can take a minute or two. After the first load, the model is cached in your browser's local storage. Every subsequent use — on the same device, same browser — is near-instant with no internet connection required.
If speed on first use matters, choose the Fast Preview engine (~5 MB). It downloads in seconds and produces a usable cut-out. The edges are slightly less refined than the larger models, but for quick drafts, mockups, or situations where you are on limited mobile data, it is the right choice.
- Fast Preview (~5 MB) — loads in seconds, good for drafts and mobile data
- General Fast (~40 MB) — balanced quality and load time
- General (~80 MB) — recommended for most photos
- Portrait (~176 MB) — best for headshots and passport photos
- Anime & Art (~176 MB) — best for illustrations and manga
2. Download format determines whether you get transparency or white fill
This is the single most common source of confusion. After background removal, users download the file and open it — sometimes seeing a white background, sometimes seeing a checkerboard pattern, sometimes seeing the background filled with white instead of transparent.
PNG and WebP both support transparency. When you open a PNG or WebP with a transparent background in most image viewers, the viewer renders the transparent areas as white or a gray checkerboard pattern — that checkerboard is the preview of transparency, not a real background. Place the file in Canva, Figma, PowerPoint, or any website and the transparent areas behave correctly.
JPG does not support transparency. If you select JPG as the output format, the removed background fills with white. This is intentional and useful for passport photos, Amazon listings, and anywhere a white background is required — but it means you cannot place the subject on a different color later without removing the background again.
Rule of thumb: download as PNG or WebP when you want to keep working with the image. Download as JPG when you need a white background right now.
3. HEIC photos from iPhone work — they are converted first
iPhones save photos as HEIC by default. HEIC files offer better compression than JPG but are not natively decodable by most browsers when fed directly to image processing pipelines.
The Irreva Background Remover accepts HEIC and HEIF input. Before sending the image to the AI model, it automatically converts the HEIC file to JPEG in your browser using a lightweight decoder. You do not need to convert your iPhone photos before uploading.
This conversion step is transparent — you will not notice it happening. The output is still the format you selected (PNG, WebP, or JPG). The only caveat is that the conversion adds a very small amount of processing time before the AI run begins.
4. Choosing the right engine makes a significant difference
All five engines in the Irreva Background Remover remove backgrounds, but each is trained on different subject types. Using the wrong engine for your image does not always cause obvious failure — but it produces edges that are slightly less clean than the right engine would.
For product photos, animals, and most objects, use General. For people, headshots, and passport-style photos, use Portrait — it handles hair and facial edges noticeably better. For anime art, manga panels, and cartoon illustrations, the Anime & Art engine understands the stylized edges of hand-drawn subjects in a way general models do not.
The Fast Preview engine is engine-agnostic — it trades some quality for speed. Use it when you need a quick result and will refine edges in an editor afterward, or when you are on limited mobile data and just want to see whether the tool will work for your image before downloading the full model.
5. Your image never leaves your browser — here is why that matters
Cloud-based background removal services process your photos on remote servers. That means your image travels over the internet, gets stored temporarily (and sometimes permanently) on a third-party system, and is subject to that service's data handling policies.
Browser-based AI tools running via WebAssembly are different in a fundamental way. The neural network model downloads to your device. When you click Remove, the pixels are processed entirely within your browser's memory. No data is sent to any server.
This distinction matters for product photos of unreleased items, for passport and ID photos that contain facial data, for medical images, and for any situation where the contents of your photo should not leave your device. It also means the tool works offline after the initial model download — useful for handling sensitive material in low-connectivity environments.
The Irreva Background Remover is explicit about this: only the AI model weights download from a CDN. Your images stay on your device from the moment you select them to the moment you download the result.
