Skip to content
Creative Studio, runs on our GPU

Open-source licenses

ToolboxHere is built on excellent open-source software. Thank you to everyone who maintains these projects. Each name below links to the project's source; the groups follow what the software does on this site.

The site and its interface

ProjectLicense
AstroMIT
ReactMIT
Tailwind CSSMIT
Lucide iconsISC
Tabler Icons (the LinkedIn icon)MIT
Simple Icons (the outlines of the brand marks on the Share buttons)CC0-1.0
dnd kit (drag to reorder files and pages)MIT

PDF and Word documents

ProjectLicense
pdf.js (Mozilla)Apache-2.0
PDFium's JBIG2 and CCITT fax decoder (black-and-white scans in PDFs, built for pdf.js)BSD-3-Clause
OpenJPEG (JPEG 2000 pictures in PDFs, built for pdf.js)BSD-2-Clause
qcms (colour profiles of pictures in PDFs, built for pdf.js)MIT
pdf-lib (the maintained @cantoo/pdf-lib version)MIT
culori (colours, used by pdf-lib)MIT
node-html-better-parser (used by pdf-lib)MIT
html-entities (used by node-html-better-parser)MIT
tslib (used by pdf-lib)0BSD
fontkit (the maintained @cantoo/fontkit version, shapes text and puts fonts into PDFs)MIT
restructure (used by fontkit)MIT
dfa (used by fontkit)MIT
brotli.js (used by fontkit)MIT
base64-js (used by brotli.js)MIT
bidi-js (orders right-to-left and left-to-right text in Word to PDF)MIT
fflate (ZIP files, and reading and writing .docx)MIT
mp4box.js (GPAC; reads MP4 and MOV video files)BSD-3-Clause
signature_pad (drawn signatures in Fill and Sign PDF)MIT

Images, photos and QR codes

ProjectLicense
heic-toLGPL-3.0
libheif (built into heic-to)LGPL-3.0
react-easy-cropMIT
normalize-wheel (used by react-easy-crop)BSD-3-Clause
scanic (finds page edges in Scan to PDF)MIT
DocAligner page-corner model (DocsaidLab): finds the page in a photo in Scan to PDF, Image to Text, ID card to A4 and Fill and SignApache-2.0
qrcode-generatorMIT
jsQR (reads each QR code back for the scan check)Apache-2.0
zxing-wasm (runs the QR code scanner's reader in your browser)MIT
ZXing-C++ (the QR code and barcode reader inside zxing-wasm)Apache-2.0
@jsquash/webp (writes WebP stickers in Safari, which cannot write WebP itself)Apache-2.0
libwebp (Google; the WebP encoder inside @jsquash/webp)BSD-3-Clause

Text recognition and on-device AI models

A model is downloaded from this site only after you start the tool that needs it, and is then kept on your device.

ProjectLicense
ONNX Runtime Web (runs the on-device AI models)MIT
Tesseract.js and tesseract.js-coreApache-2.0
Tesseract OCR and trained language dataApache-2.0
idb-keyval (used by Tesseract.js)Apache-2.0
wasm-feature-detect (used by Tesseract.js and @jsquash/webp)Apache-2.0
bmp-js (used by Tesseract.js)MIT
zlib.js (used by Tesseract.js)MIT
regenerator-runtime (used by Tesseract.js)MIT
is-url (used by Tesseract.js)MIT
PaddleOCR models (PaddlePaddle): PP-OCRv6 text detection and recognition, PP-DocLayout-S layout and SLANet_plus table structure, for Image to Text, Scan to PDF, PDF to Word and the page finder of PDF Table to Excel in the Creative StudioApache-2.0
SCOWL English word list (Kevin Atkinson), used to restore spaces in small printSCOWL (MIT-style)
Urdu (Nastaliq) OCR model: trained for this site on text rendered in Noto Nastaliq Urdu, Gulzar, Noto Naskh Arabic, Amiri, Scheherazade New, Lateef and Harmattan (all SIL Open Font License 1.1) with Tesseract's Urdu word list (Apache-2.0)Made for ToolboxHere
MediaPipe BlazeFace model (Google), ONNX version by Keijiro TakahashiApache-2.0
MediaPipe Face Mesh model (Google), ONNX version by Keijiro TakahashiApache-2.0
PP-MattingV2 human matting model (PaddleSeg)Apache-2.0
Real-ESRGAN general x4v3 upscaling model (Xintao Wang and others), ONNX version made for this siteBSD-3-Clause
BiRefNet_lite dichotomous image segmentation model (Peng Zheng and others), ONNX version made for this site, for Remove BackgroundMIT
Depth Anything V2 Small depth model (Lihe Yang and others), 8-bit ONNX version made for this site, for Blur BackgroundApache-2.0
YuNet face detector (Shiqi Yu, OpenCV Zoo), for Blur FaceMIT
MI-GAN inpainting model (Picsart AI Research), rebuilt into ONNX for this site, for Remove Object from PhotoMIT
Whisper base, small and large-v3-turbo speech recognition models (OpenAI), for Audio to TextMIT
sherpa-onnx (its ONNX export of Whisper: base and small reworked for this site, large-v3-turbo as released)Apache-2.0
DeepFilterNet3 noise suppression model (Hendrik Schröter and others), for Remove Background NoiseMIT or Apache-2.0
Demucs: Hybrid Transformer Demucs (htdemucs) music source separation model (Meta Platforms), for the Vocal Remover and Stem SplitterMIT
Qwen3-0.6B language model (Alibaba Cloud, Qwen team; ONNX export by onnx-community), for the Grammar Checker and Text SummarizerApache-2.0
Tokenizers.js (Hugging Face), the tokenizer of the Grammar Checker and Text SummarizerApache-2.0
Kokoro-82M v1.0 text to speech model and its voices (hexgrad), from the ONNX export by onnx-community, stored more compactly for this site, for Text to SpeechApache-2.0
misaki English pronunciation dictionaries (US and UK), and the text-to-phoneme rules ported to TypeScript for Text to Speech (hexgrad)Apache-2.0
graphemes_to_phonemes_en_us and _en_gb (PeterReid): the small network that guesses words missing from the dictionariesApache-2.0
lamejs (an MP3 encoder ported from LAME; the @breezystack/lamejs package), makes the MP3 in Text to SpeechLGPL-3.0
Bergamot translator (browsermt/bergamot-translator, its prebuilt WebAssembly build, unmodified), runs the translation models in your browserMPL-2.0
Firefox Translations models (Mozilla): English to Urdu, English to Arabic, Arabic to English and English to Hindi, unmodifiedMPL-2.0

Example files in the AI tools

The "Try an example" files in the AI tools, so you can see a tool work before you use your own. Photos were made smaller for this site.

ProjectLicense
Remove Background: "Woman with bindi smiles" by Oriol Casas (Unsplash, via Wikimedia Commons)CC0-1.0
White Background: "Sitting and smiling" by Brooke Cagle (Unsplash, via Wikimedia Commons), croppedCC0-1.0
Sticker Maker: "Puppy playing with a toy" by Justin Veenema (Unsplash, via Wikimedia Commons)CC0-1.0
Transparent Background Maker: a logo of a made-up bakery, drawn for this siteMIT
Blur Background: "Ergaki, Hiker in the wild, Taigish Valley, Sayan Mountains, Russia" by Vyacheslav Argenberg, resized and compressed for this siteCC BY 4.0
Blur Face: "Three friends by train tracks" by Priscilla Du Preez (Unsplash, via Wikimedia Commons)CC0-1.0
Remove Object from Photo: "Solitude" by Toa Heftiba (Unsplash, via Wikimedia Commons)CC0-1.0
Photo Enhancer: "Laughing woman in jean jacket" by Brooke Cagle (Unsplash, via Wikimedia Commons), made small and soft for the exampleCC0-1.0
Image Upscaler: "Indian man standing in doorway" by Abhas Mishra (Unsplash, via Wikimedia Commons), made 400 pixels wideCC0-1.0
Image to Text: a page of the "Fruit and Vegetable Branch administrative letter" (War Food Administration, 1944), from the Internet ArchivePublic domain (U.S. Government work)
Arabic OCR: page 1 of "PACE Code C: notice of rights and entitlements (Arabic)", Home Office. Contains public sector information licensed under the Open Government Licence v3.0.Open Government Licence v3.0
Urdu OCR: five sentences written for this site, set in Noto Nastaliq UrduMIT (the text), SIL Open Font License 1.1 (the font)
Audio to Text: "The Gettysburg Address" by Abraham Lincoln, read by John Greenman for LibriVox (the first 25 seconds)Public domain
Remove Background Noise: speech from the Rasa Urdu dataset (AI4Bharat) mixed with a vacuum cleaner recording from MS-SNSD (Microsoft)CC-BY-4.0 and MIT

Calculators, prayer times, the Qibla and readability

ProjectLicense
decimal.js (exact decimal arithmetic in the calculator, so 0.1 + 0.2 is 0.3)MIT
adhan (prayer times)MIT
GeographicLib (Charles Karney): the Qibla direction and distance on the WGS84 ellipsoidMIT
World Magnetic Model 2025 (NOAA NCEI and the British Geological Survey): corrects the Qibla compass from magnetic to true northPublic domain (U.S. Government work)
syllable (counts syllables in the Readability Checker)MIT
pluralize (used by syllable)MIT
normalize-strings (used by syllable)MIT

Fonts

Every font is served from this site, never from a font service.

ProjectLicense
Plus Jakarta Sans (the site's typeface)SIL Open Font License 1.1
Noto Nastaliq Urdu (Urdu text and the Urdu signature font)SIL Open Font License 1.1
Great Vibes (signature font)SIL Open Font License 1.1
Dancing Script (signature font)SIL Open Font License 1.1
Sacramento (signature font)SIL Open Font License 1.1
Mr Dafoe (signature font)SIL Open Font License 1.1
Aref Ruqaa (Arabic signature font)SIL Open Font License 1.1
Carlito (draws Calibri in Word to PDF)SIL Open Font License 1.1
Caladea (draws Cambria in Word to PDF)SIL Open Font License 1.1
Arimo (draws Arial in Word to PDF)SIL Open Font License 1.1
Tinos (draws Times New Roman in Word to PDF)SIL Open Font License 1.1
Cousine (draws Courier New in Word to PDF)SIL Open Font License 1.1
Gelasio (draws Georgia in Word to PDF)SIL Open Font License 1.1
Noto Naskh Arabic (Arabic text in Word to PDF)SIL Open Font License 1.1
DejaVu Sans (symbols and other alphabets in Word to PDF)Bitstream Vera Fonts licence (free), DejaVu changes public domain
Space Grotesk (invoice template font)SIL Open Font License 1.1
JetBrains Mono (invoice template font)SIL Open Font License 1.1
Newsreader (invoice template font)SIL Open Font License 1.1
Source Sans 3 (invoice template font)SIL Open Font License 1.1
Sora (invoice template font)SIL Open Font License 1.1
IBM Plex Mono (invoice template font)SIL Open Font License 1.1
Barlow Condensed (invoice template font)SIL Open Font License 1.1
Barlow (invoice template font)SIL Open Font License 1.1
DM Serif Display (invoice template font)SIL Open Font License 1.1
DM Sans (invoice template font)SIL Open Font License 1.1
Libre Caslon Text (invoice template font)SIL Open Font License 1.1
Libre Franklin (invoice template font)SIL Open Font License 1.1
Hanken Grotesk (invoice template font)SIL Open Font License 1.1
Bricolage Grotesque (invoice template font)SIL Open Font License 1.1
Cormorant Garamond (invoice template font; the supporter certificates)SIL Open Font License 1.1
Jost (invoice template font)SIL Open Font License 1.1

The Creative Studio: AI models on our own GPU

These run on our own graphics card for the paid Creative Studio, never in your browser and never for the free tools.

ProjectLicense
MiniMax H3 (video; offered outside the EU, the UK, the US and South Korea)MiniMax H3 Community License
LTX-2.5 (Lightricks, video)LTX-2.x Community License
Wan2.2 TI2V 5B (animating a picture)Apache-2.0
FLUX.2 [klein] 4B (wallpapers)Apache-2.0
Z-Image-Turbo (images with text)Apache-2.0
Qwen3-8B (slide decks)Apache-2.0
Chatterbox (Resemble AI, voice-over; adds the PerTh watermark)MIT
ACE-Step 1.5 (music)MIT
TripoSR (photo to 3D)MIT
GLM-OCR (document reader)MIT
PaddleOCR-VL-1.5 (PDF Table to Excel)Apache-2.0
Qwen3-VL-8B (questions about an image)Apache-2.0
UI-TARS-1.5-7B (screenshot helper)Apache-2.0
Qwen2.5-Omni-7B (audio, image and video questions)Apache-2.0
Freepik nsfw_image_detector (the check on every picture)MIT
Marqo nsfw-image-detection-384 (the check on every picture)Apache-2.0
Qwen3Guard-Gen-0.6B (the check on every text answer and deck)Apache-2.0
YuNet (OpenCV face detector: photos of people need consent)MIT
RapidOCR (poster titles; identity documents are refused)Apache-2.0
faster-whisper and Whisper small (checking a voice consent recording)MIT
SpeechBrain ECAPA speaker model (same voice in the consent recording and the sample)Apache-2.0

Brand marks

The marks of WhatsApp, Facebook, Instagram, Threads (Meta), Telegram, X and LinkedIn on the Share buttons are the trademarks of their owners, used only to label links to those services, as their brand guidelines allow. Using them does not mean those companies endorse ToolboxHere.

MPL software

The Bergamot translator and the Firefox Translations models are used under the Mozilla Public License 2.0, exactly as their authors publish them, as separate files that load only when you translate. Their source code is available at the links above, and the licence text sits beside the files (/vendor/bergamot/ and /models/v1/translate/).

LGPL software

heic-to and libheif are used under the LGPL-3.0 as a separate, replaceable script file that loads only when your browser cannot open HEIC photos itself. lamejs is used under the LGPL-3.0 the same way: a separate script that loads only when you save an MP3 in Text to Speech. Their source code is available at the links above.