You stay the director
The agent proposes, you decide. Every edit it makes is visible, explained and undoable in one click — direction without the busywork.
ChatGPT made writing conversational — vibe editing does it for video. Describe the edit in plain language: GPT-Video plans it, applies it through a real editing pipeline, and lets you undo any step. You direct, it executes.
Cut this podcast into 10 vertical clips with captions
The pipeline
Six steps run behind every request. Each one is a real, inspectable stage — not a single opaque "generate" button.
Drop in a file — a podcast episode, an interview, last night's stream VOD. No project setup, no timeline: the source video is the only thing you bring.
Every word is transcribed with Whisper-class speech recognition and timestamped. From here on, the whole recording is searchable text — for you and for the agent.
The engine scores candidate segments for hooks, payoffs and self-contained stories — and shows its reasoning. Every suggestion explains why it scored, so you judge with context instead of trusting a black box.
Face detection tracks the people on camera, and the vertical crop follows them. Widescreen footage becomes 9:16 without keyframing a single move.
Word-accurate captions are generated from the transcript and rendered into the video, appearing in sync with the audio — the style short-form feeds reward.
Each clip renders as a finished vertical 9:16 MP4, captions burned in, ready for TikTok, Reels and Shorts. Want a change first? Say it in chat and it's applied — and undoable.
Why it works
The agent proposes, you decide. Every edit it makes is visible, explained and undoable in one click — direction without the busywork.
Describe the edit once instead of executing it step by step. The mechanical work — scrubbing, cropping, captioning — becomes a review pass.
Clip suggestions come with their reasons and every change is a named, reversible step. You always know what happened to your video and why.
Yes. The agent never publishes or overwrites anything on its own: it proposes clips with the reasoning behind each one, applies the changes you ask for, and every step can be undone in one click.
The slowest parts of short-form editing — finding moments in hours of footage, reframing, captioning — run automatically. Your time goes to judgment calls: pick the clips, tweak what matters, post.
Clips render through a real video pipeline: face-tracked vertical reframing, word-accurate animated captions and clean encoded exports. Chat is the interface, not the renderer.
Vertical 9:16 MP4 with captions rendered in — the native format for TikTok, Instagram Reels and YouTube Shorts. Viral formats like fake text and split screens render the same way.
Create an account and try it on your own footage.
Plans from $19 a month. Cancel any day, keep the month.