System requirements
FreeClips does its work on your computer, so what it can do depends on your browser and device. Check your device below before downloading any model.
Summary
These are starting targets we are testing against, not guaranteed minimums. They will be replaced with measured results as we test more devices (see tested devices).
| Task | Starting target | What matters most |
|---|---|---|
| Editing and converting (trim, resize, merge, speed, reverse, compress, convert, GIF, audio, mute, music, silence removal) | Recent desktop Chrome or Edge; 8 GB RAM; free disk space of at least twice the output size | Whether your browser can decode your video's codec and encode the output. Long outputs are written to temporary browser storage as they're made. |
| Speech (AI clip maker, subtitles, video/audio to text) | 8 GB RAM with the fast model; 16 GB recommended for the standard model and long recordings | CPU speed (speech runs on the CPU) and free memory. A one-time model download is needed. |
| Advanced AI highlights | WebGPU with 16-bit float support; about 1 GB of graphics memory for the compact model, 3–4 GB for the large one | Your graphics adapter. Browsers can't report free graphics memory, so the large model may still fail on smaller GPUs — the compact one and the rule-based picks remain. |
| AI upscaler | WebGPU strongly recommended | Every frame is processed by the model; without WebGPU it falls back to the CPU and is very slow. The non-AI option always works. |
| Background remover | Recent desktop browser; GPU helps | A small model runs once per frame; it uses the GPU when available. |
| Phones and tablets | Experimental, no promise | Memory limits, tabs being paused in the background and codec support can stop a job from finishing. |
Browser
Use a current desktop version of Chrome or Edge. The tools need WebCodecs (to decode and encode video), Web Workers and WebAssembly with SIMD; the AI highlights and upscaler also use WebGPU. Safari and Firefox support parts of this; the device check tells you what works in yours. We check features directly rather than guessing from the browser's name.
Memory (RAM)
Browsers only give websites a rough, rounded memory hint (often capped at 8 GB), and some give none, so the check cannot tell you exactly how much RAM you have. If a job fails or the tab crashes, close other tabs, choose a smaller model or try a shorter video.
Processor and graphics
Speech and editing run on the CPU, so a faster processor means faster results. The language model for AI highlights and the AI upscaler run on the graphics card through WebGPU. If your device lacks a GPU feature, every other tool is unaffected.
Disk space
Space is used for:
- AI models, downloaded once and only when you use the tool that needs them: speech about 80 MB (standard) or about 44 MB (fast) plus a 27 MB engine; AI highlights about 580 MB (compact) or about 2.8 GB (large); speaker labels about 8 MB; background remover under 1 MB; upscaler about 2–5 MB. Sizes come from our pinned model manifest.
- Temporary files while you export, removed automatically 24 hours after you last used them (you can change this or clear them on the storage page).
- Your downloads, saved by your browser's normal download.
- Saved projects (transcripts and clip settings only) — small.
Internet connection
You need a connection to load the site and, once per model, to download it. After that, the processing itself runs locally. We have not yet tested fully offline use, so we don't promise it.
Export formats
Videos are exported as MP4 (H.264 video, AAC audio) when your browser can encode it. If it cannot, the tools offer WebM instead and say so — never labelling a WebM file as MP4. Other formats are listed only after a live check of what your browser can encode.
If something doesn't work
- Update your browser, or switch to desktop Chrome or Edge.
- Choose a smaller model, or a shorter video.
- Free up disk space, or clear temporary files and unused models on the storage page.
- In the clip maker, skip the AI highlights — the rule-based suggestions and manual clips don't need them.
Tested devices
Results from real runs. Untested combinations may still work; they are just not verified yet.
| Date | Device | Browser | What was tested | Result |
|---|---|---|---|---|
| 2026-11-20 | MacBook Pro, Apple M2 Max, 96 GB RAM, macOS 27.2 | Chrome 158 | All 18 tools with synthetic test media (53 s – 30 min 720p H.264/AAC, rotated phone MOV with mono audio, video without audio, two-voice dialogue, a public-domain portrait for background removal) | All tools export files that decode cleanly in ffmpeg, with sound aligned to the source (0 ms measured offset). Fast speech model: 53 s in 11 s (with download), 30 min in 4 min 44 s. AI highlights: compact model (580 MB) in about a minute including the download; large model (2.8 GB) in under 2 minutes, 9 s of thinking. AI upscaler on WebGPU: 2 s of 640×360 video to 720p in a few seconds. |
| This is a very fast machine, so it says little about typical laptops. Windows laptops (8 GB and 16 GB), an older CPU-only desktop, Safari, Firefox and mobile browsers have not been tested yet. | ||||