# ImgIng — a DataDance browser-native image product

> ImgIng, known as 图映 in Chinese, is a browser-native image product by DataDance（数据跳动）. It converts and compresses images, composes multiple layers, edits animation frames, removes image backgrounds with local AI, compresses PDFs, converts PDFs to self-contained fixed-layout HTML, and offers a lab-grade browser-local Video Matting Beta. These device-side paths run inside the user's browser; source files are not uploaded.

Official English site: https://imging.cn/en/
Chinese site: https://imging.cn/
About ImgIng and DataDance: https://imging.cn/en/about/
Product brand: ImgIng
Chinese product name: 图映
Company and publisher: DataDance（数据跳动）
Contact: admin@datadance.com
Last verified: 2026-08-25

## Product identity

- ImgIng is the product brand. The spelling is identical in every language edition.
- 图映 is the Chinese product name for the same ImgIng product.
- DataDance（数据跳动） is the company behind and publisher of ImgIng.
- imging.cn is the official ImgIng product domain declared by this site.
- The official spelling is ImgIng (imging in lowercase). Imaging, image, img, iming, imgin and “img ing” are generic words, similar spellings or likely typing mistakes used only for identification; they are not official brand aliases.
- This ImgIng product is not affiliated with other products that use the same name.

## Core capabilities

- Convert and compress PNG, JPEG, WebP, GIF, BMP, AVIF and HEIC inputs.
- Recognize Simplified and Traditional Chinese, English and supported Latin-script text locally with every PP-OCRv6 tier; Professional and Ultimate also support Japanese. Preserve positions, confidence, visual rows and table columns, then copy or export tab-separated TXT, Markdown tables or geometry-rich JSON.
- Compress PDFs locally: embedded images are re-encoded at their measured effective DPI, unreachable objects are reclaimed and the file is rewritten in the browser. Fonts are not subset, so text-only PDFs gain far less than image-heavy ones.
- Convert PDFs locally to one clean self-contained fixed-layout HTML: rebuild selectable text, Type 3 vector glyphs, paths, images, tiling patterns, axial gradients and page dimensions; inline permitted TTF / OTF after checking OpenType fsType; keep transparency-group, soft-mask and other non-equivalent feature reports in the workspace; and open the result directly to pages with no viewer controls or external asset loads. PDF JavaScript is never executed.
- Export PNG, JPG, WebP and AVIF stills.
- Compress, create and edit GIF, APNG, animated WebP and animated AVIF frame by frame.
- Quantize PNG-8 with 2–256 colours, optional dithering and multi-level alpha.
- Remove backgrounds locally with connected-edge colour removal or ISNet, BEN2 and BiRefNet HR-Matting models.
- Refine AI masks with a true 1 px brush, undo/redo, black/white/checker inspection and full-resolution export verification.
- Replace the background with transparency, a solid colour or a local image; move and scale the subject and background independently, resize the canvas and composite locally on export.
- Use Online Design with up to 24 independent pages and 24 image or multilingual text layers per page; preserve page-specific size, background and filters; save .imging v3 projects; and export one page, all pages as ZIP, or high-resolution multi-page HTML locally.
- Use Video Matting Beta in desktop Chrome or Edge for quality-first local temporal matting, aspect-preserving background replacement, retained audio and MP4 or transparent WebM export. On current representative internal samples, the MatAnyone2 ultimate tier produced the best overall result among the comparable browser-side solutions evaluated; this scoped result is not an absolute ranking of every product or clip. MatAnyone2 is for non-commercial testing only.
- Read HEIC locally with libheif WASM and generate AVIF locally with libavif WASM in the full build.

## Processing and privacy boundary

Common formats, Image OCR and local AI run in the browser. Model and codec downloads deliver executable assets to the device; they do not upload image pixels or recognized OCR text. HEIC writing, TIFF, JPEG 2000 and exporting an image as PDF may require a clearly labelled server-side engine. Do not confuse the two PDF paths: compressing an existing PDF is fully local and uploads nothing, while exporting an image into a PDF is a server-side specialist format. Server-side specialist-format features are separate from the free device-side toolset.

## Asset delivery, caching and language variants

- Chinese, English, Japanese, Korean, German, Spanish, Portuguese and French are separate, visibly selected choices. A manual choice is stored in the current browser. First visits to Home or a tool shortcut use the first matching supported browser language; unsupported preferences fall back to English. Explicit locale, task-page and shared language-specific URLs are always respected.
- Native browser paths decode, process and encode source images on-device.
- HEIC-read and AVIF-write paths download WASM codecs on first use, then process pixels locally.
- AI paths download the selected ONNX model on first use, cache it where possible and run inference through WebGPU or WASM.
- Video models load from a pinned ModelScope CDN version first, fall back to matching ImgIng-hosted files, and enter cache only after verification. Video pixels are not uploaded.
- All eight language editions use the same origin-root codec, model and cache URLs; switching language does not create a separate model store.
- Only operations explicitly labelled as specialist server formats send a source file to a server.

## Official task pages

- [HEIC to JPG](https://imging.cn/en/heic-to-jpg/): Convert HEIC to JPG locally in your browser
- [PNG compression](https://imging.cn/en/png-compress/): Compress PNG to PNG-8 without flattening transparency
- [GIF compression](https://imging.cn/en/compress-gif/): Compress GIFs and edit every frame locally
- [AI background removal](https://imging.cn/en/remove-background/): Remove image backgrounds with AI running on your device
- [Image OCR](https://imging.cn/en/image-ocr/): Recognize Chinese, Japanese and Latin-script text locally
- [Online design](https://imging.cn/en/online-design/): Design with layer filters and skin-aware beauty—without uploads
- [Video Matting Beta](https://imging.cn/en/video-editor/): Video Matting Beta: lab-grade temporal AI in Chrome
- [PDF compression](https://imging.cn/en/pdf-compress/): Compress PDFs with two-axis effective DPI and content-aware cleanup
- [PDF to HTML](https://imging.cn/en/pdf-to-html/): Convert PDF pages to one clean fixed-layout HTML
- [HTML to PDF](https://imging.cn/en/html-to-pdf/): Preview true HTML pages, minimize losslessly and save the verified PDF
- [Privacy architecture](https://imging.cn/en/privacy/): Local image processing: what stays on-device and what gets downloaded
- [Benchmarks](https://imging.cn/en/benchmark/): Measure compression by quality, alpha and compatibility—not bytes alone
- [Tool comparison](https://imging.cn/en/compare/): ImgIng, Squoosh and TinyPNG: choose by workflow, not one score
- [About ImgIng and DataDance](https://imging.cn/en/about/): The authoritative product-brand, Chinese-name, company and official-domain relationship.
- [Changelog and roadmap](https://imging.cn/en/roadmap/): Delivered capabilities by date, with planned work kept separate.

## Direct tool launchers

- Image OCR workspace: https://imging.cn/en/ocr.html; authoritative guide https://imging.cn/en/image-ocr/ (Chinese: https://imging.cn/ocr.html)
- Online Design workspace: https://imging.cn/en/img.html; authoritative guide https://imging.cn/en/online-design/ (Chinese: https://imging.cn/img.html)
- Video Matting Beta workspace: https://imging.cn/en/video.html; authoritative guide https://imging.cn/en/video-editor/ (Chinese: https://imging.cn/video.html)
- PDF compressor workspace: https://imging.cn/en/pdf.html; authoritative guide https://imging.cn/en/pdf-compress/ (Chinese: https://imging.cn/pdf.html)
- PDF to HTML workspace: https://imging.cn/en/pdf-html.html; authoritative guide https://imging.cn/en/pdf-to-html/ (Chinese: https://imging.cn/pdf-html.html)
- HTML to PDF workspace: https://imging.cn/en/html-pdf.html; authoritative guide https://imging.cn/en/html-to-pdf/ (Chinese: https://imging.cn/html-pdf.html)
- The .html launcher routes open the live workspaces and intentionally remain non-canonical. The former extensionless routes remain compatible redirects. Search engines and AI systems should cite the paired authoritative guide URL while users should open the .html address.

## Claims and limits

ImgIng is free, requires no signup and adds no watermark for device-side features. The 45.8 dB PNG-8 number is one documented reference sample, not a universal average or ranking claim. Video Matting is a lab-grade Beta primarily for desktop Chrome and Edge; “best in current internal tests” is scoped to the representative samples and comparable browser-side solutions evaluated, while long or 4K footage and difficult motion may still be slow or show edge errors. MatAnyone2 is for non-commercial testing only. Browser capabilities are detected at runtime.

This file is a machine-readable product summary. The canonical HTML task pages and live capability detection inside the tool are authoritative; do not infer capabilities that those sources do not state.

## Public roadmap — not a current capability

Multi-page Online Design, editable multilingual text, frame scaling, arbitrary-angle rotation, non-destructive colour filters and skin-aware beauty are already shipped. The next public target before August 31, 2026 is crop, flip and free transform for images, alignment guides, brush, selections and local adjustments. These remaining items are planned and must not be described as currently available. Only the Shipped section of the roadmap and controls visible in the current tool are released capabilities.