Free AI tools
AI models that download once and then run inside your browser: remove backgrounds and objects from photos, enhance them, read text from images, transcribe audio and read text aloud with natural voices. You get the convenience of AI without handing your files to a cloud service.
AI Creative Studio
Video with sound, wallpapers, thumbnails with titles, slide decks and voice-overs: 13 AI tools too heavy for a browser. Runs on our GPU
Write what you want, see the price and the wait, and download the result.
A paid section, made on our GPU. The free tools on this site stay free and run in your browser.
See the studioEvery tool in this category is free, needs no account and runs on your device. New tools are added as people ask for them; suggest one.
About the AI tools
The AI tools run machine-learning models on your own device. Each page says how large the model is before anything is downloaded, downloads it only when you tap the button, and keeps it in the browser’s cache so the next visit is instant. The photos, recordings and scans you give a model stay on the device, and you can check by switching the network off once the model has loaded. The models are open-source releases with their licences listed on the licences page: PaddleOCR’s PP-OCRv6 for printed English text, Tesseract for Arabic, our own recogniser trained for Urdu Nastaliq, OpenAI’s Whisper for speech, Kokoro for reading text aloud, Real-ESRGAN for photo enhancement, BiRefNet for cutting the background out of photos, MI-GAN for removing objects from photos and DeepFilterNet for noise.
Running on the device means the models are smaller than the ones behind cloud services, and the pages say so: Image to Text reads clean printed text well and struggles with handwriting; Audio to Text offers a fast model and a more accurate one that takes longer, and the Urdu page measures its accuracy against a held-out set rather than claiming perfection. Where a model did not clearly beat the alternative it replaced, it was not shipped. The pages state measured figures, not marketing ones.
The tools on this page run in your browser. The heavier models (video with sound, AI wallpapers, slide decks, narrated voice-overs) run on our own GPU in the Creative Studio, a paid section of the site.
Frequently asked questions
How big are the downloads?
From a few megabytes for the OCR and noise models to about 70 MB for the fast speech model, about 105 MB for the text to speech voices and several hundred megabytes for the most accurate speech model. Each page shows the size before you start, and the download happens once; after that the model loads from the browser’s cache.
Why is the AI slower than an online service?
It runs on your processor instead of on a server farm, with one thread in most browsers. A minute of speech takes about twenty seconds on a laptop with the fast model. The trade is privacy: the recording never leaves the device.
Does it work on a phone?
The OCR, noise and photo tools do. The speech tools work on recent phones with the fast model; the most accurate speech model needs a computer with 8 GB of memory and is only offered there. Text to Speech works on phones too, more slowly than on a computer, and shows the time left as it goes.
Is the Arabic and Urdu recognition reliable?
Printed Arabic reads well with Tesseract. Urdu printed in Nastaliq needed a model of our own, because general engines get about a quarter of the characters wrong; the Urdu OCR page shows its measured error rate and lets you correct doubtful words in place.
Which tool do I need?
- Text out of a screenshot, a scan or a photo of a page
- Image to Text, Arabic OCR or Urdu OCR
- A voice note, lecture or meeting as text
- Audio to Text
- A noisy recording
- Remove Background Noise
- Text read aloud by a natural voice, or an MP3 of it
- Text to Speech
- A small or blurry photo that must be larger
- Photo Enhancer or Image Upscaler
- A photo without its background, on white, or as a WhatsApp sticker
- Remove Background, White Background or Sticker Maker; Transparent Background Maker for logos and signatures
- Someone or something to take out of a photo
- Object Remover