Help Center All FAQs in one place

It depends on the mode: quality-based compression only does gentle lossy optimization with almost no visible difference; compressing to a target KB lowers quality first, then resolution—size first.

No. Compression runs locally in your browser; images never leave your device.

You can drop up to 300 images for batch processing; download each one or pack them all into a ZIP.

The default 80 has minimal visible loss for JPG; for print or quality-sensitive work use 90+ or lossless mode.

Up to 300 images, 50MB each, all processed locally in the browser—no queue.

Only in your browser memory; download individually or as a ZIP—the server keeps nothing.

It finds the highest quality not exceeding the target, usually slightly under (e.g. 99.6KB) so any validator passes.

JPG, PNG and WebP; PNG→JPG drops size a lot and the page notes the format change.

For extreme targets (e.g. 10KB) also reduce dimensions; the tool prompts when downscaling is required.

JPG is lossy by nature; defaults show minimal loss—use PNG lossless mode for true lossless.

No—only encoding quality changes; pixel size stays unless you use target-size mode and allow downscaling.

Yes—any JPG up to 50MB.

Yes. Lossless only cleans redundant data and optimizes encoding—pixels and alpha unchanged.

No—alpha is always kept; lossy only affects color quantization.

JPG has no transparency and fills it black—keep PNG instead.

Mainstream browsers and WeChat/Alipay mini-programs support it; convert back to JPG for legacy IE.

Yes, WebP by default; you can also choose JPG or PNG in settings.

Yes—frame count, loop and aspect ratio stay the same.

Uploads are deleted right after processing; results are kept 2 hours then cleaned up.

No—SVGO is lossless; only if you disable viewBox and the image relies on it might display change.

Yes—up to 100 SVG files at once.

Yes—only metadata is removed, no re-encode; pixels stay as-is and can be hash-verified.

Because it keeps quality and only clears redundancy; use lossy or target-size for major savings.

Flattened by default; imported subfolders keep their structure.

Not much—JPG/PNG are already compressed, so re-deflating gains little and slows packing; the UI states which was used.

Files are named image_01, image_02 in order; extensions unchanged.

No—duplicates become file(2).jpg, so both survive extraction.

No—file size barely changes; print size changes with DPI.

96 is the Windows screen default; this tool can rewrite it to any target.

Resize only changes pixel dimensions with unchanged quality; compression aims to shrink file size and may adjust both quality and dimensions.

You can adjust the watermark's position, opacity and size; by default it sits bottom-right, and you can also tile it across the whole image.

Yes. Select several images, set a uniform width/height or percentage, export in batch, or download as a ZIP.

Enlarging always interpolates—prefer shrinking; use the detail-preserving algorithm when you must enlarge.

Yes, batch mode outputs all images at the same target size.

No—the original stays local; a new file is exported.

Yes—drag a box or type x/y/width/height in source pixels; both sync live.

No blank area—dragging is bounded and typed values are clamped to the image.

Blur flattens everything including edges; denoise cross-checks similar patches to keep edges and texture.

PNG if the source has alpha, else JPG (quality 92)—lossless PNG would balloon after denoise.

Process twice: strength 10 first, then add a bit—preserves more detail than one pass at 30.

Watermarks are baked into pixels—keep the original; semi-transparent marks stay visible without blocking the image.

Yes, rendered with system Chinese fonts—no install or upload; size is a ratio of the short edge, consistent across sizes.

PNG (transparent, best), JPG and WebP; the logo is read locally and never uploaded.

JPG can't hold transparency; the tool fills white by default, and you can pick another color.

Lossy targets (JPG/WebP) encode at your quality; PNG/BMP are lossless.

Re-encoded at your quality (default 95) with no visible difference; pixel size stays original.

Chrome/Edge on Windows can't decode HEIC without a local decoder or a server; this tool decodes server-side and deletes the original immediately.

No—EXIF is stripped on re-encode so location doesn't leak publicly.

Yes, multi-select and queue; each file uses 1 quota—20/day free, more on VIP.

Transparent areas fill with the background color, or crop to content.

Pixels are untouched—only metadata and encoding redundancy change.

Unlimited locally, bounded by browser memory; keep to ~20 per run, huge images may lag.

No—each scales proportionally by the short edge; grid mode centers at original ratio with padding.

Enter a color code: #ffffff white, #f5f5f5 light gray; blank means transparent.

No—stitching runs entirely in the browser; images never touch the server.

Yes—each piece is an exact equal division of the original.

Yes—set columns and rows to 3; up to 10×10.

Many, packed into a ZIP named part_01, part_02 in order.

No. Splitting and packaging happen locally in your browser.

Yes—the PNG alpha is written into the ICO.

No login needed—free users get 20 images/day. After logging in you can view history; upgrade to VIP for 200–1000/day.

Not long-term. The original is deleted right after processing; result files are kept 2 hours for download.

Model inference uses server compute, so quota controls cost; local tools like compress and resize use no compute and stay free and unlimited.

Free users get 20/day, no login needed; upgrade to VIP for more.

No. Uploads are deleted after processing and never enter any training set.

VIP supports up to 50 images per batch.

Only the background changes—face and size untouched—and output is standard JPG, accepted by application systems.

Collar edges get refined; use Fine mode, and brighten the original first if needed.

Yes—front-facing, no hat, even lighting; the tool auto-crops and corrects.

Choose print layout to get a 6-inch photo-paper layout a studio can print directly.

No redrawing of features—it is overall denoise and sharpening; Strong is just a bigger boost.

Up to 20MB each—fine for phone originals and typical scans.

Semantic inference is accurate for large areas like sky and plants; small-object colors may differ from reality.

98%+ on clear printed text; handwriting and very blurry images score lower—sharpen first.

Uploads are deleted immediately after recognition—never stored.

JPG→PNG is lossless with no quality loss; PNG→JPG is lossy and fills transparent areas white. For WebP you choose the quality.

HEIC is Apple's proprietary format browsers can't read. Use the HEIC to JPG tool: the server decodes it to JPG or PNG, and the original is deleted immediately after.

JPG has no transparency; transparent areas get filled. To keep transparency convert to PNG or WebP, or fill the background white first.

Photo PNGs shrink a lot as JPG; small solid-color images may grow slightly—normal.

Transparent by default; choose White under Background if needed.

No—SVG is vector; scaling re-renders with sharp edges.

Uses width/height if present, else viewBox; if neither, outputs 1024×1024.

No—SVG renders to PNG in the browser; files never touch the server.

No—PNG only stores current pixels; detail lost to JPG cannot be restored.

All modern browsers do; keep a JPG copy for legacy environments.

Yes—split into JPGs per page and packed as a ZIP.

Covers major brands from the last decade; unsupported formats are clearly flagged.

AVIF is high-fidelity; JPG at default 95 looks identical.

No. Merge, split, compress and image-to-PDF all run locally in your browser; document content never leaves your computer.

A scan is essentially an image; run OCR to extract text first, then lay it out and export as needed.

Yes. Split and merge only move page data and do not re-compress page content.

Text stays vector and unaffected; images resample by level—Standard is barely distinguishable.

Yes—drag list items to reorder.

Open-password PDFs aren't supported yet; permission-only passwords convert fine.

Free tier allows up to 20 files / 500 pages per task.

By original page range, e.g. report_p1-5.pdf, for easy identification.

Yes—choose a page range; others stay clean.

You can clip any segment, but keep it under 10 seconds—longer means bigger files, and chat/social apps also cap GIF size.

Lower the frame rate, shrink dimensions and reduce colors all cut size noticeably; for static scenes keep only keyframes.

GIF supports at most 256 colors, so gradients and shadows lose detail. For high quality send an MP4 or WebP animation instead.

Free tier allows up to 100 frames per task.

Under 100 frames usually finish within 10 seconds.

Barely—frame content is unchanged, only order and duration reverse.

Output as-is, with a note that a single-frame image looks the same reversed.

Minor lossy re-encode (MP4/MOV CRF 23, near-lossless; WebM VP9 CRF 32, smaller at same quality).

No—only container and codec change; duration, resolution and content stay.

MP4, MOV, WebM, MKV and more; up to 200MB per file.

Outputs standard MP4/H.264, compatible with WeChat, QQ and email.

Usually a pure white background (RGB 255,255,255), the subject fully visible and uncropped, taking over 70% of the frame; check each platform's category rules.

Yes. Background swap, resize and watermark all support multi-select batch processing, then download as a ZIP.

Pick the size template for your target platform, or use the resize tool to batch-export multiple specs.

Output is pure white and square, meeting mainstream main-image rules; follow platform rules for special categories.

The product model is tuned for transparent and reflective materials; Fine mode refines edges.

Up to 12 per run—upload in batches. Each file uses 1 quota: 20/day free, more on VIP.

Packed as a ZIP, numbered by upload order (001, 002…), so order survives extraction.

Only that one is skipped; the rest process and appear in the ZIP.

Use the single White Background tool per image—it lets you pick red, blue, gray, etc.

Default Cover keeps proportions; Contain adds white padding.

Yes—all images use the same position and opacity for a consistent look.

Yes—switch to Image watermark and pick the logo; it scales to the chosen position, all locally in the browser.

Compression, resizing, watermarking, format conversion, cropping, lossless compression and DPI changes all run in your browser—images never touch the server. Only features that need a model or a special decoder (AI cutout, ID photos, photo restoration, HEIC to JPG) run server-side; the original is deleted right after processing, and results are kept from 2 hours up to 7 days depending on your plan.

No. Compression, editing, conversion and PDF tools work without signing up. Registering is only useful when you want to check your quota or task history.

Local tools are unlimited. AI features include 20 images per day for free users, rising to 200–1,000 on VIP plans. Quotas reset at 00:00 daily and do not roll over.

Local tools handle up to 300 images at once; server tasks are processed one by one from a queue, and each file uses one run from your quota.

2 hours for free users, and 24 hours / 72 hours / 7 days on VIP plans. Expired files are removed by a cleanup job.

No. Files processed server-side are isolated by task number—only someone holding the task number can download them, and they are deleted when the task expires.

Yes. You can download again at any time while the task is retained, or adjust the settings and submit once more—another AI run is deducted.

Pick a plan on the VIP page, scan the Alipay / WeChat QR code and pay; the quota is credited as soon as the payment succeeds. If your bank was charged but the page never activated, send us the order number and we will credit it after verification.

VIP is a virtual service, so the days already used are non-refundable. For unused time, duplicate payments, or payments that never activated, send the order number to support; after verification we refund the unused part and roll back the membership period and that day's quota.

Submit your account on the "Reset password" page; once verified, support will reset it. The new password is shown only once, so keep an eye on it.

Try another browser first (the latest Chrome or Edge is recommended). If it still fails, send the page address and what happened to "Contact us" and we will follow it up as a ticket.