Convert Files — free file converter, nothing uploaded

Live at convertfiles.in — a free online file converter and a set of file tools that run entirely in the browser. Nothing is uploaded: the file is read, converted and downloaded on the visitor's own device.

• 399 conversions, each with its own page — all of them. Popular: HEIC to JPG, PDF to Word, Word to PDF, JPG to PDF, PNG to JPG, MOV to MP4, MP4 to MP3, WebP to PNG, CSV to Excel, STL to GLB.
• 21 free tools — all of them. Compress image to a size (20–500 KB for forms), Compress PDF, Passport photo maker, Merge PDF, Split PDF, Sign PDF, Image to text (OCR), Resize image, Trim video, MP3 cutter, Remove EXIF, QR code generator, Unzip online.
• Guides in English and Hindi.
• Embed it on your own site with one line: the widget.

How it works technically: Canvas and WebCodecs for images, audio and video; pdf.js, pdf-lib and jsPDF for PDFs; Tesseract (WebAssembly) for OCR; hand-written encoders for GIF, BMP, ICO, WAV, TAR and glTF. No server touches a file.

Going live (convertfiles.in on GitHub Pages)

Everything the public site needs is generated by one command from site.json:

python3 build.py

It stamps the AdSense / analytics tags into every page, versions the assets, writes privacy.html, terms.html, contact.html, 404.html, one page per popular conversion (/heic-to-jpg/), one per guide (/guides/<slug>/), sitemap.xml, robots.txt, ads.txt and CNAME. Run it after any edit, then commit and push — GitHub Pages publishes the main branch as-is.

site.json holds the only settings you should ever need to change:

key	what it is
domain	https://convertfiles.in — used for canonical URLs, the sitemap and CNAME
contact_email	shown on the Contact page and in the legal pages
adsense_client	your AdSense publisher ID, ca-pub-…; blank = no ads
adsense_slots	optional ad-unit IDs for the result / home / article placements; blank = Auto ads only
ga4_measurement_id	optional Google Analytics 4 id, G-…; blank = no analytics

AdSense, step by step: sign up at adsense.google.com with the live domain → paste the ca-pub-… id into site.json → python3 build.py → commit and push → click "Verify" in AdSense (the tag is now in every page's head and ads.txt is published) → after approval, turn on Auto ads; if you also create manual ad units, put their slot ids in adsense_slots.

DNS at your registrar (one-time): four A records on the apex (@) pointing to 185.199.108.153, 185.199.109.153, 185.199.110.153, 185.199.111.153, and a CNAME record www → <your-github-user>.github.io. Then in the GitHub repo → Settings → Pages tick "Enforce HTTPS" once the certificate is issued (up to an hour after DNS propagates).

Run it

python3 tools/file-converter/serve.py

Then open http://localhost:8777. serve.py is the standard library server with caching switched off, so edits show on a plain reload; python3 -m http.server 8777 --directory tools/file-converter works as well. (There is also a .claude/launch.json entry named convert-files.) A plain file:// open mostly works too, but PDF rendering needs the page served over HTTP.

When a conversion finishes the result card appears in place, with a preview and a Download button; the same button sits on every card in Your files.

An internet connection is needed on first use: conversion libraries (pdf.js, jsPDF, SheetJS, JSZip, mammoth, marked, js-yaml, UTIF, lamejs) are pulled from CDN on demand and cached by the browser. Nothing else leaves the machine.

What it converts

About 1,600 conversion pairs across 81 source formats in Chrome. The exact number depends on the browser — at load the app probes for WebP encoding, CompressionStream and the WebCodecs AAC/Opus/H.264/VP9 encoders, and removes any target the browser cannot actually produce rather than offering an option that is guaranteed to fail. The "Formats" tab lists the live graph.

Both format pickers are searchable: type an extension (docx), a format name (spreadsheet) or a plain word (word, excel, photo, iphone, quicktime). The From list shows how many targets each format has.

Group	In	Out
Images	png jpg webp gif bmp avif jxl svg ico tif heic heif ppm pgm pbm	png jpg webp avif jxl bmp gif ico icns svg tif tga ppm pgm pbm · pdf · docx odt html epub rtf md (picture embedded)
Documents	txt log md html docx odt rtf pdf epub	txt md html pdf docx odt rtf epub tex · png jpg webp (page renders)
Spreadsheets	xlsx xls xlsm xlsb ods fods csv tsv dif dbf prn slk	xlsx xls xlsm xlsb ods fods csv tsv dif dbf prn slk html rtf · pdf docx odt epub tex (real tables) · json jsonl yaml xml toml sql md txt
Data	json jsonl ndjson yaml yml xml toml	json jsonl yaml xml toml sql · every spreadsheet format · txt md html pdf docx odt tex
Audio	mp3 wav m4a aac ogg oga opus flac weba aiff	wav mp3 m4a aac ogg opus
Video	mp4 m4v webm mov ogv	mp4 (H.264 + AAC) · webm (VP9 + Opus) — a real re-encode · gif · png jpg webp avif jxl bmp tga (single frame) · wav mp3 m4a aac ogg opus (audio track)
Archives	zip tar gz tgz	zip tar gz tgz
3D / AR	obj stl ply gltf glb	obj stl ply gltf glb
Fonts	ttf otf woff	woff · ttf otf
Anything else	any file	zip tar gz tgz

Every document pair goes through one shared model (js/docmodel.js) that carries headings, bold/italic runs, links, lists, quotes, code, tables and embedded images, so a Word file with a picture and a table still has both after becoming PDF, ODT, EPUB, RTF or HTML.

Audio and video re-encoding use the browser's own WebCodecs encoders; the containers are written by mp4-muxer / webm-muxer, except Ogg Opus, ADTS, WOFF, GIF, BMP, TGA, ICNS, Netpbm and TAR, which are written by hand in js/encoders.js and js/convert.js.

Size caps

Set per family in js/formats.js, sized for what a browser tab can hold: images 80 MB, documents 60 MB, spreadsheets 60 MB, data 120 MB, audio 250 MB, 3D 250 MB, fonts 40 MB, video 600 MB, archives 800 MB. Video re-encoding is additionally limited to 15 minutes of footage, and output is capped at 1920 px wide unless a width is set. The picker shows the cap and blocks anything over it.

What it cannot do

• Video re-encoding runs at roughly real time or faster for 1080p on a recent laptop, because every frame is decoded by the <video> element and re-encoded by WebCodecs; a ten-minute clip takes a few minutes. Keep the tab open — Chrome suspends video decoding in background tabs.
• HEIC and JPEG XL need a one-time download outside Safari (1.3 MB and ~1 MB WebAssembly decoders); AVIF/JXL encoding likewise.
• WOFF2 cannot be unpacked — it uses Brotli, which browsers expose for fetches but not to page scripts.
• .ai, .psd, .sketch and friends are not inputs. They are proprietary containers with no browser decoder.
• Raster to SVG is not a trace; it embeds the bitmap inside an SVG wrapper.
• PDF text extraction is text only. A scanned PDF has no text layer.
• 3D conversions carry geometry only. Materials, textures, animations and Draco-compressed meshes are not carried across; USDZ is not written.

Several files at once

Pick as many files as you like (drop, browse, or "Add more"). With one file the From tile can be overridden; with several, each file keeps its own type and the To list is the intersection of what every type can become (mixed PNG + MD

• CSV still share PDF, DOCX, HTML, EPUB, ODT, RTF and the archive formats). One tap converts them all in sequence with a running "3 of 8" status; the result card lists every file with its own Download plus Download all as .zip, and failures are shown per file without stopping the batch. Every output is stored in Your files and counted once. Up to 200 files per batch.

On phones (coarse pointer) the wording switches to "Choose files", a sticky Convert bar follows the page once files are chosen, and the format picker opens as a bottom sheet.

Site content

Below the converter the home page carries the editorial material a public site needs, all rendered from js/content.js in the same visual system:

• At a glance — live numbers: conversions available and formats in/out (computed from the graph in that browser), files converted worldwide (the real total from GET /public on the analytics API, cached a minute at the edge, with this device's share as the subtitle), and "0 files uploaded". If the API cannot be reached the tile falls back to this device's own count.
• How it works — three steps.
• Popular conversions — 32 curated pairs; each tile presets From/To and opens the file picker. Pairs the browser cannot do are hidden.
• Your most used — the pairs converted most on this device.
• Guides — seven short articles (HEIC, PDF↔Word, image formats, MOV→MP4, 3D→GLB, privacy, CSV vs Excel) with a "try it" button. Linkable as #guide:<slug>; the list is #guides.
• About and a nine-question FAQ.
• Support — a UPI tip section: the QR (assets/upi-qr.png, cleaned to pure black/white so it scans and weighs 1.3 KB), the UPI ID with a Copy button, and on phones an "Open in UPI app" button (upi://pay deep link). After a conversion the result card carries a quiet "Leave a tip" link. Below it sits a single sponsor slot: fill SPONSOR in js/content.js (name, url, tagline) and it renders "Sponsored by …"; leave it empty and it shows the open slot. SUPPORT.upiId / payeeName / qr live in the same file.

There is no sign-in, sign-up, pricing or account anywhere; the site is free and public by design.

Admin panel, usage counts and the Android app

• convertfiles.in/admin — sign in with the admin password. Dashboard: views, unique visitors and conversions for today / 7 / 30 / 90 days, a daily chart, top conversions, top pages, referrers, devices, failed conversions. Content: editors for every file in content/ plus site.json; "Save & publish" commits through the API and the site rebuilds itself (GitHub Actions, about a minute). Countries and states/regions come from what Cloudflare resolves at its edge (the IP is never read or stored, only a daily count per place), and average time on the site from the seconds each page was actually visible.
• api/ — a Cloudflare Worker with a D1 database, deployed at https://convertfiles-api.convertfiles.workers.dev. It stores daily counters only (see api/schema.sql): no files, no names, no IPs, no identifiers. The site sends a view event on page load and a convert / fail event after a conversion, and skips both when the browser sends Do Not Track. Deploy with api/deploy.sh; secrets with api/set-secrets.sh.
• tools/convertfiles_admin — the Flutter app "Convert Files Admin" (Android). Sign in with the same password; same dashboard, pull to refresh. Build: flutter build apk --release --split-per-abi.

Your files

Every conversion is kept in the browser (IndexedDB database format-bench — the internal name is unchanged so files converted earlier stay visible — store conversions, keyed by an auto-increment id, indexed on createdAt and targetExt; each record holds the output Blob plus name, source, sizes, MIME type, timing, options and timestamp).

The Your files tab shows them as a gallery — a thumbnail for images, a format badge for everything else — with Download and Delete on every card, a search box, and the browser's storage usage. Clicking any card opens a preview inside the page (browsers block window.open on blob URLs, so a new tab is not an option): images, PDF, audio, video and text formats render directly, and formats with no browser viewer — Word, Excel, ODT, EPUB, RTF — are rendered by converting them to HTML in memory with the same engine, shown in a sandboxed frame. Anything genuinely unviewable says so and offers Download. The raw record is deliberately not shown in the UI; open the browser's DevTools → Application → IndexedDB if you ever need to see it.

Layout

site.json           domain, contact email, AdSense and analytics ids
build.py            generates the production pages (see "Going live")
pages/              templates for the legal pages
index.html          markup, option controls, the format picker
serve.py            local dev server (no-cache headers)
css/app.css         theme tokens, light and dark
js/formats.js       extension metadata, family size caps, lazy CDN loader
js/encoders.js      hand-written binary writers: GIF (median cut + LZW), BMP, ICO, ICNS, TGA, Netpbm, WAV, TAR, Ogg Opus
js/docmodel.js      the shared document model: 8 parsers, 9 writers, images and tables
js/convert.js       the conversion graph, one rule per from-set -> to-set
js/db.js            IndexedDB wrapper
js/content.js       popular conversions, guides, FAQ and about copy
js/app.js           UI
test/run.sh         headless tests

Tests

tools/file-converter/test/run.sh

145 checks covering the parts no library provides: the GIF encoder is verified by decoding its own output with an independent LZW decoder (exact round trip on indexed colour, bounded error on gradients, animation delays, the NETSCAPE loop block, transparency), plus BMP row padding and alpha compositing, WAV headers and sample scaling, TAR checksums and round trips, OBJ/STL/PLY parsing and writing, glTF building and parsing (including node transforms), GLB packing, Ogg page CRCs and lacing, TGA/Netpbm/ICNS layout, the SQL writer and the sfnt/WOFF table directory. Runs on macOS JavaScriptCore via osascript, so it needs no node or npm.