
Free AI tools that run in your browser
Background Remover
On your deviceCuts the person or object out of a photo and gives you a transparent PNG or a new background colour.Model: IMG.LY ISNet (fp16)First use downloads about 100 MB, then cachedRuns on your device (CPU, WebAssembly)AI Image Upscaler
On your deviceEnlarges a small or blurry image 2× or 4× and adds believable detail, instead of just stretching pixels.Model: Real-ESRGAN general x4v3First use downloads about 29 MB, then cachedRuns on your device: WebGPU when available, otherwise CPUAI Photo Colorizer
On your deviceAdds plausible colour to black-and-white and sepia photos, with a colour-strength slider.Model: Mini Photo Colorizer v3First use downloads about 40 MB, then cachedRuns on your device: WebGPU when available, otherwise CPUVocal Remover
On your deviceSplits a song into a vocal track and an instrumental track for karaoke, practice or remixes.Model: Spleeter 2-stem (Deezer)First use downloads about 63 MB, then cachedRuns on your device: WebGPU when available, otherwise CPUAudio Noise Remover
On your deviceRemoves fan hum, hiss, traffic and room noise from voice recordings and the audio of videos.Model: RNNoise (Xiph.Org)First use downloads about 4.8 MB, then cachedRuns on your device (CPU, WebAssembly)Image to Text (OCR)
On your deviceReads the text in screenshots, scans and photos (OCR) in 14 languages, ready to copy.Model: Tesseract 5 (LSTM)First use downloads about 7.1 MB, then cachedRuns on your device (CPU, WebAssembly)AI Background Blur
On your deviceSegment a subject with IMG.LY, blur the background locally and export an opaque PNG. Radius changes reuse the cached subject; inspect hair and edges.Model: IMG.LY ISNet (fp16)First use downloads about 100 MB, then cachedRuns on your device (CPU, WebAssembly)Private Translator
On your deviceTranslates text between 39 languages and keeps Markdown, code and links intact.Model: Chrome's built-in translatorChrome downloads the language model itself (size not reported)Runs on your device, in Chrome's built-in modelNeeds Chrome 138 or later on a computer
These free AI tools run a real AI model on your own computer or phone, inside the browser tab. Remove a photo background, upscale or colorize an image, split a song into vocals and music, clean noise out of a recording, read text from a screenshot, or translate privately. No account, no watermark, no daily limit.
Your file is never uploaded: the model comes to your device once, and the work happens there. Each card names the model, its licence and how much the first use downloads, so you know exactly what you are running.
Why is this free? (What's the catch?)
Many online AI tools send your file to a server with a powerful graphics card (GPU). Every run costs the provider money, so they ask for an account, add a watermark, sell credits or set a daily limit. That is a fair way to pay for a GPU server, and it is why people expect a catch.
Here, the model runs on your own device instead. There is no GPU server doing the work, so a run costs us nothing extra, and there is nothing to ration or charge for. The site itself is paid for by supporters, not by your files or your data.
The honest costs are on your side and small: the first use downloads the model (from about 5 MB to about 100 MB, listed on each card), which your browser then keeps for next time; and speed depends on your device. A recent computer is faster than an old phone, and the tools that can use WebGPU (your graphics chip) are much faster with it.
Models, licences and download sizes
Sizes are what the first use downloads (model plus its engine, where the engine is a separate file), measured on 2026-11-14. After that, your browser keeps them.
| Tool | Model | Licence | First download | Runs on |
|---|---|---|---|---|
| Background Remover | IMG.LY ISNet (fp16) | AGPL-3.0 | 100 MB | CPU (WebAssembly) |
| AI Image Upscaler | Real-ESRGAN general x4v3 | BSD-3-Clause | 29 MB | WebGPU or CPU |
| AI Photo Colorizer | Mini Photo Colorizer v3 | Apache-2.0 | 40 MB | WebGPU or CPU |
| Vocal Remover | Spleeter 2-stem (Deezer) | MIT | 63 MB | WebGPU or CPU |
| Audio Noise Remover | RNNoise (Xiph.Org) | BSD-3-Clause | 4.8 MB | CPU (WebAssembly) |
| Image to Text (OCR) | Tesseract 5 (LSTM) | Apache-2.0 | 7.1 MB | CPU (WebAssembly) |
| AI Background Blur | IMG.LY ISNet (fp16) | AGPL-3.0 | 100 MB | CPU (WebAssembly) |
| Private Translator | Chrome's built-in translator | Google Chrome terms | Set by Chrome | Chrome's built-in model |
A typical online AI tool compared with these
Written in general terms: plenty of online services differ, and some are excellent. The difference comes from where the model runs.
| Compared | Typical online AI tool | These tools |
|---|---|---|
| Account | Often needed to download, or to get more uses | None |
| Your file | Uploaded to the provider's servers and processed there | Stays on your device |
| Watermark | Common on free plans | None |
| Limits | Daily caps or credits, because every run uses a paid GPU | No daily limit; only a size or length limit per file, so your device does not run out of memory |
| How it is paid for | Subscriptions, credits or ads | Supporters pay for the site; a run costs nothing extra |
| Speed | Fast on any device, because the server's GPU does the work | Depends on your device; a one-time model download first |
| Model size | Can use very large models | Smaller models that fit in a browser: very good at one job, not a general chatbot |
Honest limits
- The first use downloads the model. On a phone or a slow connection, start on Wi-Fi; the sizes are listed above.
- Big files need memory. An old phone may be slow or run out of memory on a long song or a very large photo; a computer handles them better.
- The translator only works in Chrome 138 or later on a computer, because it uses the model built into Chrome. We do not fall back to an online service.
- These are focused models, each trained for one task. They are good at that task, but they are not a general AI assistant.
Questions
Are these AI tools really free?
Yes. There is no account, no trial, no credits and no watermark. They can be free because the AI model runs on your own device, so we do not pay for a GPU server each time you use one.
Is my photo, song or text uploaded anywhere?
No. The model is downloaded to your browser and your file is processed there. You can check it in your browser's developer tools (Network tab): no request carries your file.
Why does a tool download something the first time?
That is the AI model itself. It ranges from about 5 MB (noise remover) to about 100 MB (background remover). Your browser keeps it, so the next use starts much faster and does not download it again.
Do I need a graphics card or a fast computer?
No. Every tool works without a dedicated graphics card. A newer computer is faster, and the upscaler, colorizer and vocal remover use WebGPU (your graphics chip) when the browser offers it, which can turn minutes into seconds.
Do they work on a phone?
Most do, in a current mobile browser, but more slowly than on a computer, and very large files can run out of memory. The translator needs desktop Chrome.
Which AI models do you use?
Open-source and licensed models, named on each card: IMG.LY ISNet for backgrounds, Real-ESRGAN for upscaling, Mini Photo Colorizer v3, Deezer's Spleeter for vocals, Xiph.Org's RNNoise for noise, Tesseract 5 for OCR, and Chrome's built-in translator.