Free · No sign-up · On your device
Private AI Studio
Free AI tools that run on your device. Transcribe audio, caption video, translate, summarize a PDF or describe your images, and your files never leave your phone or computer.
- Runs on your device
- Files never uploaded
- No account
- Free
Privacy
Your files stay with you
Most online AI tools upload your file to a server. These do not. Here is what that means in practice.
Runs on your device
The AI model loads into your browser tab and does the work on your own processor.
Files never leave it
Audio, video, PDFs, text and images are not uploaded. There is no server copy to leak or keep.
Download once, keep it
The model is saved in your browser. You download it the first time and reuse it after that.
Works without a connection
Once a model is saved, the work itself needs no internet. Only the first download does.
The tools
Five tools, one privacy promise
Every tool is free and needs no account.
Audio to text
Drop a recording and get a timestamped transcript. Whisper runs in this browser tab, so your audio never leaves your device.
Whisper speech model · MP3, M4A, WAV or a recording. Also gives SRT and VTT.Video to SRT
Drop a video and get captions you can edit and load into YouTube, Premiere or CapCut. Everything runs on your device.
Whisper speech model · MP4, MOV or WebM. Edit timing, then export.Translate
Translate a paragraph or a subtitle file without sending it anywhere. Timestamps and line breaks are kept.
Firefox Translations models · English to and from Hindi, Spanish, German, Chinese and more.Summarize PDF
Get the key points of a long PDF without uploading it to a chatbot. Each point links to its page so you can check it.
pdf.js reading, instant summary · Works on text PDFs. Scanned pages need OCR first.Alt text generator
Get alt text and a caption for your images without uploading them. Edit the suggestion, then copy it or download a CSV.
Florence-2 vision model · Up to 20 images, with a CSV download.
How it works
From file to result in four steps
Pick a tool
Choose audio, video, translation, PDF or images.
Download the model once
Each tool shows the size first, then saves the model in your browser.
Drop in your file
It is processed inside the tab. Nothing is uploaded.
Copy or download
Get text, SRT, VTT, CSV or Markdown, ready to use.
Devices and browsers
What you need
Everything runs on your own hardware, so speed depends on your device.
Computers
- A recent Chrome, Edge, Firefox or Safari.
- Audio and video up to 2 hours per file.
- WebGPU, where available, makes the larger models practical.
iPhone, iPad and Android
- Safari or Chrome on iPhone, Chrome on Android.
- Audio and video up to 30 minutes per file; alt text handles 5 images at a time without WebGPU.
- If a device runs out of memory, the tool says so instead of failing silently.
Storage and languages
- Models are saved in your browser's storage; each tool shows the size and has a delete button.
- English is strongest. Hindi transcription and Hindi PDF summaries are beta.
- Private browsing may clear the saved model when you close the window.
FAQ
Questions people ask
Are my files uploaded anywhere?
No. The AI models run inside your browser tab, so your audio, video, PDFs, text and images stay on your device. This website only serves the page and the model files, the same way it serves any page.
Does it work offline?
Partly. After a model has downloaded and been saved in your browser, the work itself needs no connection. You still need to load the page, and the first run needs the internet to download the model.
Why does the first run download so much?
The model is the AI. It has to be on your device for your files to stay there. Each tool shows the download size before it starts, and the model is saved so you only download it once.
Is it free? Do I need an account?
It is free and there is no sign-up. The tools run on your hardware, so there is nothing for us to charge for.
Which browsers and devices work?
A recent Chrome, Edge, Firefox or Safari on a computer, and Safari or Chrome on an iPhone, or Chrome on Android. Where the browser supports WebGPU the tools use it for speed; otherwise they fall back to WebAssembly, which is slower.
How accurate are the results?
Good for drafts and everyday use. English is the strongest language. Hindi transcription and Hindi PDF summaries are in beta and can be wrong, so read the result before you rely on it. Summaries and alt text can make mistakes.
How do I remove the models from my device?
Open any tool and use the delete button on its model card. That frees the space the model uses in your browser.
Try one now
Pick a tool, drop in a file and see it work without leaving your device.