Home › How Mshopi works
How Mshopi works
Mshopi is built so your photos never leave your device. Here is exactly what runs where, which open-source AI models are used and under which licences, and how you can check it yourself.
Everything runs in your browser
Most online editors upload your files to a server, process them there and send the result back. Mshopi does the work on your own computer or phone, inside the browser tab:
- Video: your photos and clips are decoded and composed on an HTML canvas, then encoded to MP4 with the browser's built-in WebCodecs encoders (H.264 video and AAC audio, with VP9 and Opus as a fallback on browsers without H.264). Your graphics hardware does the heavy lifting.
- Audio: mixing, fades, noise reduction, sound effects and the generated music use the Web Audio API.
- Prompt reading: the one-line prompt is read by rules in the page itself (prices, currencies, discounts, phone numbers, delivery words, occasions, languages). No text is sent to an AI service.
- Storage: projects, imported files, cut-outs, generated music and your brand kit are stored in your browser (IndexedDB). Nothing is stored on our servers.
Open AI models and their licences
Three optional features use open AI models that run in your browser through the open-source Transformers.js library. Each model downloads once from Hugging Face and is then cached by your browser. We chose models whose licences allow commercial use, because your videos are for selling.
| Feature | Model | Licence |
|---|---|---|
| AI background removal | BiRefNet lite | MIT |
| Automatic captions | Whisper base (OpenAI) | Apache 2.0 |
| AI voice-over (English) | SpeechT5 (Microsoft) | MIT |
The plain-background remover, the music generator and every other feature use no AI model and no download.
What is downloaded, and from where
- From mshopi.com: the app itself (a few hundred kilobytes), icons and pages.
- From jsDelivr (cdn.jsdelivr.net): the MP4 writing library when you export, and the Transformers.js library when you first use an AI feature.
- From Hugging Face (huggingface.co): AI model files, only when you first use that feature.
- From Google Fonts: the fonts used in the editor and in your videos.
These services see a normal download request from your device, like any website. Your photos, videos, prompts and recordings are never sent to them or to us.
Check it yourself
Open your browser's developer tools (F12 on a computer), go to the Network tab and make a video. You'll see downloads of the app, fonts and libraries, but no upload of your photos.
Installable app
Mshopi can be installed as an app on Android, Windows, macOS and ChromeOS from Chrome or Edge (look for the Install app button or the install icon in the address bar). The installed app always loads the newest version from the internet.
Browser support
Mshopi needs a modern browser with WebCodecs for exporting video: Chrome, Edge, Safari 17 or later, or Firefox 130 or later, on Windows, macOS, Linux, ChromeOS, Android or iPhone. A computer is faster for AI features and long videos.
Who is behind it
Mshopi is made by Muhammad Naeem, an independent developer in Pakistan. It is built on the same engine as Lumnexvideoeditor. Questions are welcome on the contact page; see also the privacy policy and what's new.
Frequently asked questions
Does Mshopi upload my photos?
No. Photos and videos are processed in your browser on your own device and are never uploaded.
Which AI models does Mshopi use?
BiRefNet lite (MIT) for background removal, Whisper base (Apache 2.0) for captions and SpeechT5 (MIT) for the English voice-over, all running in your browser.
Does Mshopi need an internet connection?
Yes, to open the site, export videos and download AI models the first time. Your photos are still processed on your own device and never uploaded.
Make your product video now
Mshopi is free, needs no account and never adds a watermark. Your photos stay on your device.