Video to Prompt Generator

What a video to prompt generator does

A video to prompt generator watches a clip and writes down what it sees as text: what is in frame, how the camera moves, how the scene is lit and how long each shot lasts. You can paste that text into a video model such as Sora 2, Veo 3 or Kling to make something similar, or keep it as a shot list for your own edit. ShotPrompt does this for clips of up to 60 seconds.

It is upload only. You choose a file from your device, so TikTok, Instagram and YouTube links are not supported. Download or screen-record a clip you have the right to use, then upload that file.

Who it is for

Creators who found a shot they want to study, editors who need a shot list from a rough cut, and anyone learning how a scene was filmed. If you can describe what you want but not name the camera move, the output gives you the words.

How it works

1. Choose a video file. MP4, MOV and WebM usually work, up to 100 MB and 60 seconds.
2. Pick the target model: Sora 2, Veo 3, Kling, Seedance, Runway or Generic.
3. Tick the two confirmations and press Analyze video. Your browser takes one still frame per second and scales it so the longer side is 512 pixels.
4. Only those frames, with their timestamps, are sent to our server. An AI model (Gemma on Cloudflare Workers AI) describes them in order. The video file never leaves your device.
5. Read the result, switch between the three formats and copy the one you need.

A 20-second clip becomes about 20 frames. Because every frame carries its time, the description can put a start and an end on each shot instead of treating the clip as one blur. Camera movement is inferred from how the frames change, so a push-in or a pan is an informed reading, not a measurement. Check it against your clip.

Three output formats from one run

• Shots: one card per shot with a time range such as 0:00–0:04, a description, and camera and lighting notes. Use it as a shot list or storyboard, or to rebuild a clip one shot at a time.
• Full prompt: one block of text written for the model you picked. Paste it straight into the generator.
• JSON: the same content as structured fields, handy for API workflows, spreadsheets and tools that take structured prompts.

Each tab has its own Copy button, and a style line at the top sums up the look of the whole clip. If you mostly want structured output, open video to JSON prompt. For model-specific notes, see Sora prompts from video and Veo 3 prompts from video.

What is different here

• Shot by shot, with time codes. The result is a list of timed shots, not one paragraph about the whole clip.
• Six targets, three formats. Choose Sora 2, Veo 3, Kling, Seedance, Runway or Generic and get cards, a full prompt and JSON in one run.
• Frames only. The video file stays on your device. We send still frames, and we do not store the video or the frames.
• No sign-up. You get 3 free analyses per day, with no account.

What it cannot do

• It reads frames only, so sound, speech and music are not analyzed. Add audio notes yourself.
• It samples one frame per second, so a cut shorter than a second can be missed.
• The output is an AI-generated description. It can be wrong, and no prompt will reproduce a clip exactly in any video model.
• It is instructed not to identify real people in a video.
• Clips over 60 seconds or 100 MB are not accepted.

Privacy in one line

Your video file is not uploaded; only still frames are sent to an AI model (Gemma on Cloudflare) to write the description, and we do not store the video or the frames. Details are on the Privacy page.