Open-source notices
InfoHubs uses the following open source software: File conversion takes place in the user's browser.
- @ffmpeg/ffmpeg 0.12.15 — MIT. Browser conversion interface.
- @ffmpeg/core 0.12.10 — Build WebAssembly, including FFmpeg and codec libraries. This core, which contains LGPL/GPL components and includes libx264, is covered by the GPL. Build recipe and source reference, FFmpeg License.
- heic-to 1.3.0 — MIT. HEIC/HEIF decoding. Includes libheif 1.20.2 and libde265; Please refer to the corresponding source project for component licenses. libheif source.
- fflate 0.8.2 — MIT. Create ZIP.
- rhwp / @rhwp/core 0.8.7 — MIT. HWP·HWPX parser, typesetting/rendering, editing document export. Full license text.
- pdf-lib 1.17.1 — MIT. Create PDF of images. Full license text.
- opentype.js 1.3.4 — MIT. Analysis of user-linked fonts. Full license text.
- Transformers.js 3.8.1 —Apache-2.0. Local voice recognition. Full license text.
- ONNX Runtime Web 1.22.0-dev.20250409-89f8206ba4 — MIT. Single-threaded WASM inference. Full license text.
- Xenova/whisper-tiny —Apache-2.0. ONNX conversion of OpenAI Whisper Tiny multilingual model. Model revision 5332fcc35e32a33b86612b9a57a89be7906102b1, q8.
- Xenova/whisper-small — Apache-2.0. ONNX conversion of OpenAI Whisper Small multilingual model. Model revision 2d67713f236afa48a18992566e7647f6ca848e13, q8.
- PDF.js 5.6.205 (legacy browser build) — Apache-2.0. License.
- Tesseract.js 6.0.1 / Tesseract.js-core 6.0.0 / tessdata 4.0.0 — Apache-2.0. License.
- SheetJS CE 0.20.3 — Apache-2.0. License.
- qrcode 1.5.4 / jsQR 1.4.0 — MIT / Apache-2.0. QR generation license, QR reader license.
- jsdiff 8.0.2 — BSD-3-Clause. License.
- Noto Sans KR Regular (static weight 400 derived from the variable font) — SIL OFL 1.1. License. Korean/English caption font.
- U²-NetP saliency model — Apache-2.0. License. Model source: U-2-Net; ONNX artifact: rembg model release. Runs locally and may produce incorrect edges.
- 7z-wasm 1.2.0 / 7-Zip 24.09 — GNU LGPL 2.1+ and unRAR restrictions. Large compressed files are handled by the device. license, unRAR limitations, WASM build source/recipe, 7-Zip original source.
- hash-wasm 4.12.0 — MIT. In-device piece-wise hash/HMAC calculation. license.
- @pdf-lib/fontkit 1.1.1 — MIT. Korean PDF fonts included. License/Source.
- LightCraft — MIT selection applied. Tone curve, sRGB conversion, and display LUT have been separated into small local WASM modules. license, NOTICE.
- PdfCraft — MIT selection applied. Use the COS·Filter·Encryption submodules. license, NOTICE. Browser memory/object count limits and verification adapters have been added. Edit/Source, Dependencies/License List.
- Life·Developer Additional Tools — Fixed version browser bundle with character encoding, SQL·XML·YAML·Cron, and GIF/ICO processing. Dependency·Version·License·SHA-256. Include the full text of the component licenses in that folder.
- OpenCV Zoo YuNet face candidate model — MIT. license, Model source·Fixed commit·SHA-256. Face detection runs on the local WASM.
external connection
The conversion engine and libraries are loaded from this site. Automatic caption models are downloaded from Hugging Face and may remain in your browser cache. The contents of the files selected by the user will not be transferred. Large unzipped results may remain in your device's temporary storage and can be removed by clearing results or deleting browser site data. You can use it as a system font even if there is no font connection.
scope of support
We do not guarantee all file formats or codecs. Check the main screen for supported formats and restrictions. Camera RAW, PSD and DRM protected files are not supported. HEIC/HEIF·AVIF·SVG in the general converter is for input only. A separate SVG tool will also output an SVG copy containing only the elements it supports.
MIT License
Copyright (c) ffmpeg.wasm contributors, heic-to contributors, and Arjun Barrett (fflate).
Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions: The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software. THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
Local AI models
Automatic face crop reuses the existing YuNet model. Object selection runs SlimSAM and speech detection runs Silero VAD on your device. Model sources, revisions and licenses are listed below.
Additional local AI models
Real-ESRGAN, EfficientNet-Lite0, PP-LCNet, RNNoise and fastText run on your device. Original files and text are not uploaded. Model sources, hashes and licenses are listed below.
- Real-ESRGAN — Xintao Wang — BSD-3-Clause
- Swin2SR — mv-lab — Apache-2.0
- PP-LCNet — PaddlePaddle — Apache-2.0
- EfficientNet-Lite0 — TensorFlow / ONNX Model Zoo — Apache-2.0
- RNNoise — Xiph.Org — BSD-3-Clause
- RNNoise WASM — Jitsi — Apache-2.0
- fastText — Facebook / Meta — MIT
- fasttext.wasm.js — MIT
- fastText lid.176.ftz — Facebook / Meta — CC-BY-SA-3.0
- ImageNet labels — TorchVision — BSD-3-Clause
- Models — provenance.json
PaddleOCR
PaddleOCR.js 0.4.2 · PP-OCRv5 mobile detection + Korean recognition. ONNX Runtime Web 1.24.3 / OpenCV.js run on your device. OCR images are not uploaded.
MediaPipe / MoveNet local models
EfficientDet-Lite0, YAMNet, FaceLandmarker, GestureRecognizer and MoveNet Lightning run on your device. Files are not uploaded. Model and runtime sources, versions and hashes are recorded below.