The subtitle generator writes captions from speech and burns them in, and remove silence from video cuts pauses and “um”s in one pass. The AI clip maker (Beta) finds highlights and turns them into 9:16 shorts, auto reframe video follows faces when you go vertical, and the video background remover (Beta) needs no green screen.

Describe an edit, blur faces, bleep words

With the AI video editor (Beta) you type what you want, such as “cut the pauses, add captions, make it vertical”, approve the plan and it runs the steps on your device. Blur face detects and tracks faces in photos and videos, and the bleep censor finds swear words and covers them with a beep or silence.

Transcripts, chapters and translation

Audio to text and video to text produce TXT, SRT or VTT with timestamps, and WhatsApp audio to text handles voice notes. The YouTube chapters generator turns a transcript into 00:00 timestamps, translate subtitles keeps every cue time, and the YouTube description generator (Beta) drafts a title, description and hashtags.

AI audio and photo tools

For sound, remove background noise takes hiss, fan noise and mains hum out of speech, even in a video, and text to speech turns a script into MP3 or WAV with American or British English voices. On the photo side, the AI models handle cut-outs with remove background, enlargement with the image upscaler, retouching with remove object from photo and OCR with image to text.

How on-device AI works

Most tools download an open model the first time you use them, after you confirm the size, and your browser caches it; a few small models are built in. Depending on the model and your device, it runs on your graphics chip through WebGPU or on the processor through WebAssembly. Your files are read in the page and never sent to an AI service.

What each AI model downloads, once
Model	Used for	Download
Whisper Tiny, Base or Small	Captions, transcripts, chapters, auto cut, bleeps	44 MB, 80 MB or 252 MB on the CPU; 65 MB, 113 MB or 325 MB with WebGPU
Whisper Large v3 Turbo	Most accurate captions (WebGPU only)	566 MB
MODNet	Fast background removal, video backgrounds	7 MB
BiRefNet Lite or ormbg	Quality background removal	44 MB to 115 MB, depending on your GPU
Swin2SR	Image upscaler	53 MB (Photos & JPGs) or 54 MB (Clean art & renders)
MI-GAN	Remove objects from photos	28 MB
Kokoro	Text to speech	93 MB
OPUS-MT	Subtitle translation when Chrome’s built-in translator is unavailable	107 MB to 137 MB per language direction
MiniLM	Topic changes for English chapters	24 MB
RNNoise	Noise reducer	5 MB, part of the site
YuNet	Face blur, auto reframe, AI clips, passport photos	233 KB, part of the site

Sizes are the decimal MB the download dialog shows. GPUs without 16-bit float support get larger Whisper files (Turbo: 762 MB). Text recognition loads a small language pack instead.

AI tools FAQ

What is on-device AI?

On-device AI runs the model on your own computer or phone instead of a company’s server. The page downloads the model once, then your browser does the work, so audio, video and photos stay with you and no account or API key is needed.

Are free AI tools safe for private files?

Cloud AI tools have to receive your file to process it. These do not: the model comes to your device, and the file never leaves it. That makes them suitable for interviews, client footage, ID photos and other material you would rather not upload.

Which AI tools work without sign-up?

All of them. There is no account, no credit system and no watermark on results. The tools marked Beta work, but expect rough edges on difficult footage.

Can AI tools run offline?

Yes, once the model is cached. Use a tool online once so the page and its model are stored, and it keeps working in airplane mode. Clearing your browser’s site data removes the models, and they download again next time.

Explore other categories

All tools

Free AI Tools for Video Editing, Audio & Photos

AI tools · 17

Quick conversions & presets · 3

Also handy for ai · 1