About AudioVideoClips

What this site is

AudioVideoClips is a collection of small, single-purpose media utilities: compress an image, cut a video, extract audio from MP4, transcribe speech, remove silence, generate a voiceover. Every tool is free, needs no account, and shows no watermark. The whole catalogue is listed on the homepage.

The guiding constraint is privacy by architecture. There is no upload queue, no server-side processing and no file storage. Your media is decoded, edited and encoded by code running on your own CPU, and the result is handed back to you as a download.

How the on-device stack works

Most classic media tools are thin wrappers around a server that does the real work, which means uploading your footage and waiting in a queue. We took the opposite route:

• WebAssembly for encoding and decoding. FFmpeg is compiled to WebAssembly and runs in your tab, so codec work — trimming, re-encoding, format conversion — happens locally.
• On-device AI for understanding. Speech recognition uses Whisper-class models, and the fast mode uses the 16.9 MBWhistlemodel on the Needleengine by Cactus Compute. Both are Apache-2.0 licensed and documented in ourthird-party notices.
• One-time downloads, then offline. Engines and models are fetched once from public CDNs and model hubs, cached by the browser, and reused afterwards. The app shell works offline via a service worker.
• A full editor alongside the micro-tools. Thefull editor adds a timeline, captions, beat sync, denoise, text overlays and text-to-speech voices for longer jobs.

Who it is for

Creators and marketers trimming social clips, journalists and students transcribing interviews, developers converting media formats, and anyone who simply does not want to hand a private recording to a third-party server. If a tool does not need a server, we do not build one.

How it is funded

The tools are free to use with no account and no usage cap. The site is supported by a small amount of non-intrusive advertising; ads never sit on top of a tool's controls, and no tool is gated behind an ad. We do not sell data, and we do not build profiles of visitors.

Limitations worth knowing

Browser-side processing is not magic. Long files take proportionally longer than short ones because the work competes with the tabs you have open; AI models need a one-time download; and the accuracy of a 16.9 MB speech model is lower than a 145 MB one on noisy or accented audio. Where a heavier model gives a better result, the tool says so and lets you choose.

Get in touch

Bug reports, feature requests and corrections are welcome on thecontact page. For how your data is handled, read the privacy policy; for the rules of using the site, see the terms of service.