Upload a clip or paste a link. Get the transcript, structure, scene breakdown and prompts you can reuse in any AI video model.
Create a reusable prompt from video
A video to prompt generator watches a video and writes the text prompt that would recreate it with an AI video model. Video2Prompt does this scene by scene: it describes the subject, action, framing, camera movement, lighting and style of each shot, and gives you one prompt for the whole video.
Upload a clip or paste a TikTok, Instagram Reels, YouTube, Douyin, Xiaohongshu or Kuaishou link. The AI watches and listens, then writes the prompts.
Long videos don't fit in one prompt. You get a prompt for every scene plus a full-video prompt, so each shot can be regenerated on its own.
Pick a prompt style — Seedance 2.x, MiniMax H3, Kling or Runway — and the wording follows what that model responds to best.
From a reference video to prompts you can paste into any AI video tool.
Drop an MP4, MOV or WebM file (up to 30MB and 5 minutes) or paste a public video link.
Pick the AI video model you want to use and the output language, or let it follow the video's language.
Copy all prompts, export the scenes as CSV, or click Generate video to render a scene with Prompt to Video.
A full breakdown of how the video is made, so the prompt actually reproduces it.
A ready prompt for each scene and one for the full video, in the style of your target model.
Subject, action, shot size, camera movement, lighting and color for every scene.
Every spoken line with speakers and timestamps, plus important on-screen text.
The hook, main beats and CTA, with a script quality score and the reasons behind it.
The overall visual style, pacing and mood of the video, summarized for reuse.
Copy everything at once or export the scene list as CSV for your team.
Whatever you call it — video to prompt, prompt from video or a video prompt generator — these are the jobs it does.
Recreate a reference video with AI: turn every shot into a prompt you can render in your favorite model.
Keep the structure and camera work of a video you like, but swap the subject, product or setting.
Start from a trending video instead of a blank page. Edit the generated prompts into something new.
Need a still? Each scene description works as an image prompt for a key frame or thumbnail.
Use the full-video prompt with an image or video reference to guide motion and style in models that support references.
See how an AI-generated video was likely prompted, then learn from the wording.
Choose a prompt style before you analyze. You can run the same video again in another style any time.
Concise, shot-focused prompts with clear subject, action and camera direction. Works with Seedance 2.0, 2.0 Fast, Mini and 2.5.
Prompts that describe motion and expression in detail, suited to character and performance shots.
Structured prompts with explicit camera moves and scene layout.
Short, visual prompts that lead with style and composition.
Upload the video or paste its link above, choose a prompt style and click Analyze video. In about a minute you get a prompt for every scene plus one for the whole video.
Prompt styles for Seedance 2.x, MiniMax H3, Kling and Runway. A Sora style is also available for long, cinematic prompts.
Yes. Paste a public link from TikTok, Instagram Reels, YouTube or Shorts, Douyin, Xiaohongshu or Kuaishou. Videos can be up to 5 minutes long.
Yes. Click Generate video on any scene to open Prompt to Video and render it with Seedance 2.x or MiniMax H3.
Sign up to get 5 free credits and check in daily for more. Each analysis costs 4 credits.
Yes. Each scene description works well as an image prompt for a key frame, a thumbnail or a storyboard frame.
Short-form videos with clear scenes: ads, UGC, product demos, talking heads, tutorials and AI-generated clips. Very fast montages may be grouped into fewer scenes.
Yes. Every analysis includes the transcript, hook, beats, script score and shot list along with the prompts.
Dedicated pages for the platforms you study most.