What the AI can and cannot do
Item by item: what the AI can do, what needs sign-in, what it can't.
People often say Make a 10-second birthday greeting with colorful balloons floating upCut the first 10 seconds and add a title cardAdd subtitles to this clip
Item by item: what TutuCut’s AI can do, what needs sign-in, and what it clearly can’t. Editing, subtitles, transitions, effects, motion graphics, 3D scenes, a walking, lip-synced character — and the limits: no live-action footage, no quality repair, no watermark removal, no 3D models from nothing.
This page says it plainly: what works now, what needs sign-in, and what we clearly don’t do. Better to list it up front than have you find out after trying for half an hour.
The table below is not marketing copy. “Whatever you can do by hand, the AI must be able to do too” is an automated test on this site, and the build fails if one is missing — so the table stays in step with the code.
What it does not do is written down too: no photographic live-action generation, no restoring quality that was never recorded, no stripping someone else’s watermark. It is here rather than discovered halfway through a project.
| Capability | How the AI does it | Needs internet? | How long |
|---|---|---|---|
| Trim, split, merge | Edit timeline.tsx | The AI chat needs internet; editing itself doesn’t | Seconds |
| Speed changes, freeze frames | Do clip.speed · clip.freeze | The AI chat needs internet; editing itself doesn’t | Seconds |
| Subtitles and text (you supply the words) | Write timeline.tsx | The AI chat needs internet; editing itself doesn’t | Seconds |
| 13 transitions · 20 effects · 26 filters | Edit timeline.tsx | The AI chat needs internet; editing itself doesn’t | Seconds |
| Motion graphics (shapes, paths, text) | Do scene.create-graph · scene.set-values | The AI chat needs internet; editing itself doesn’t | Seconds |
| Code-driven motion · 32 templates · Canvas / SVG / GSAP / Three.js | Do motion.add · Write /motion/<id>.html | The AI chat needs internet; editing itself doesn’t | Seconds |
| Footage in motion clips: images, videos, GIFs, fonts; video windows can carry sound | Do motion.set-asset · motion.audio · motion.fetch-external · Frame asset | The AI chat needs internet; editing itself doesn’t | Seconds |
| 3D scenes · 5 basic shapes · lights and materials · import .glb | Do scene.* | The AI chat needs internet; editing itself doesn’t | Seconds |
| A walking character · lip sync to your voice (8 mouth shapes) | Do scene.create-character · scene.bind-speech | The AI chat needs internet; editing itself doesn’t | Seconds |
| 6 templates | Do template.apply | The AI chat needs internet; editing itself doesn’t | Seconds |
| Mark the music beats | Do audio.beat-markers | The AI chat needs internet; editing itself doesn’t | Seconds |
| Import, export, undo | Import · Export · Undo | The AI chat needs internet; editing itself doesn’t | Depends on file size |
| Transcribed subtitles (what’s being said) | Listen → Subtitles | Yes — that audio is sent to speech recognition | Ten-odd seconds |
| Change only the clips, time range or frame area you picked (the AI looks at it first) | Frame area:"selection" · at:"selection" | The AI chat needs internet; editing itself doesn’t | Seconds |
What it definitely can’t do
It cannot: generate real live-action footage of people · sharpen or upscale blurry video · remove existing watermarks · identify who is in the video · create a 3D model file from nothing · reverse playback · open WMV / FLV / RM
Every way the AI can act: 16 tools
LSReadWriteEditGrepGlobUndoListenSubtitlesImportReplaceExtractExportFrameDoTask
FAQ
- What can the AI do right now?
- Editing (trim, split, merge, speed, freeze frames), subtitles and text, 13 transitions, 20 effects, 26 filters, 6 templates, motion graphics, 3D scenes (including imported .glb models), a walking character that lip-syncs to your voice, bone animation, frame-by-frame drawing, changing the aspect ratio, volume and music, and export to MP4/GIF/MP3/images. None of it needs sign-in or a queue.
- What needs sign-in?
- Two things: saving projects to the cloud, and AI features that go through TutuDrive — having the AI listen to audio to make subtitles, and placing generated images on the timeline. Everything else, including talking to the AI to change your video, needs no sign-in; the AI allowance is whatever the assistant shows.
- What definitely can’t it do?
- It does not generate live-action footage of real people, restore damaged quality, do super-resolution, remove an existing watermark, identify who is in a video, create a 3D model file from nothing (it can only import .glb), or handle WMV / FLV / RM. Without a network the AI is unavailable, but local editing carries on.
- Why won’t it do those things?
- Two kinds of reason. Some can’t reach acceptable quality in the browser alone (repairing quality or removing a watermark means rebuilding every frame); others touch other people’s rights (identifying people, removing someone else’s watermark). Better to say plainly that we don’t do them than to ship something poor that lets you down.
- Does any of this upload my video?
- No. The AI reads the project structure — which clips there are, where they sit, what they say. When it needs to see the picture it grabs a few small frames (or the area you boxed) in your browser; the video file itself is never uploaded. The four things named above need a server, but none of them uploads your video.
- Will this table be updated?
- Yes, and it is enforced. If a new capability is added without being listed here, the check fails.