Video to prompt

What a video prompt needs, shot by shot
- Framing: wide, medium, close-up, over the shoulder.
- Camera move: static, push in, pan, follow, handheld.
- Who is in frame: each person once, by name, with what they look like described separately.
- Action: what happens in this shot, in the order it happens.
- Setting, time and light: the place, day or night, the weather, where the light comes from.
- Look: the visual style, kept the same for the whole video.
- Sound and dialogue: the line spoken and who says it, or that the shot is silent.
Video to prompt: a shot sheet to fill in
| Field | What to write | Example |
|---|---|---|
| Shot | Number and length | Shot 3, about 4 seconds |
| Framing and move | Size and camera | Medium shot, slow push in |
| In frame | Names only | Mara, Theo |
| Action | What changes | Mara slides the letter across the table; Theo does not pick it up |
| Setting | Place, time, light | Kitchen at night, one lamp over the table |
| Dialogue | Who says what | Mara: “Read it.” |
Write the looks once, on a separate line per person: age, hair, clothes, one distinct detail. Then every shot only needs the names, and the people stay the same from shot to shot.
Turning a video into a prompt by hand
- Mark the cuts. Watch once with the sound off and note where each shot starts.
- One line per shot. Framing, move, who, action, setting, in that order.
- Describe the people once. A short appearance line for each, referred to by name afterwards.
- Add the dialogue last. Listen again and put each line under its shot.
A one-minute clip with fast cuts can have twenty shots or more, which is where reading it automatically saves the most time.
Video to prompt with SceneMixer's remake
On the home page, Remake a video with Remake the story takes one MP4 or MOV clip, up to 5 minutes. The read is priced by the clip's length in minutes, rounded up, and the price is on Start remake before you press it; the number of cuts does not change it.
Reading your clip runs four passes: it finds the cuts, identifies the cast, reads every shot for picture and dialogue, and writes up the scene list and lines. Here is what we found then counts the shots, characters, scenes, props and lines.
What you get is the prompt, already split: every person becomes a character with a written appearance and a voice description, every place a location, every shot a numbered beat with its framing, camera move, who is in frame and the line spoken. Edit any of it as text before anything is drawn. The source clip is used to read the story only and is never given to the video model, so the faces that come out are the ones on your own character sheets.
One prompt or a whole storyboard
For a single clip, one prompt of a few lines is enough. For anything with cuts and recurring people, a storyboard works better: the looks live on character sheets, each shot keeps its own short prompt, and changing one shot does not disturb the others.
Make it in SceneMixer, start to finish
- Upload: on the home page open Remake a video, choose Remake the story and pick one MP4 or MOV clip, up to 5 minutes. The upload starts at once and the price of the read is on the button before you commit.
- Read: press Start remake. Reading your clip finds the cuts, identifies the cast, reads every shot and writes up the scenes and lines; you can leave and it finishes without you.
- Check the prompts: Here is what we found lists the shots, characters, scenes, props and lines. In the storyboard every shot is a numbered beat with its framing, camera move, who is in frame and what is said; edit any of it as text.
- Step 1: choose the Visual Style, the Scene Style and Costume Style and a Target Region; the read tells you what happens, you decide how it looks.
- Step 2 and video: each person is drawn once as a character sheet (the first 15 setup images on an account are free), then the segments render on the default tier at 10 credits a second.
| Writing the prompt by hand | SceneMixer's remake read | |
|---|---|---|
| Input | Watching the clip | One MP4 or MOV, up to 5 minutes |
| Shots | You mark the cuts | Cuts found for you |
| People | You describe each one | Each becomes a character with an appearance and a voice |
| Dialogue | You transcribe it | Read from the clip and placed under its shot |
| Output | A text document | An editable storyboard, ready to render |
FAQ
What is video to prompt?
Turning an existing video into the text that describes it to a video model: shot by shot, who is in frame, what they do, where, the camera, the look and the dialogue.
Can AI turn a video into a prompt?
Yes. SceneMixer's Remake a video tab reads a clip shot by shot and writes it up as a storyboard with the cast, locations, framing, camera moves and lines.
How long can the video be?
Up to 5 minutes for a remake read in SceneMixer, as MP4 or MOV; the limits are shown on the upload box.
What does it cost?
SceneMixer prices the read by the clip's length in minutes, rounded up, and shows the price on the button before you start. Rendering is then priced step by step.
Does it copy the original video?
No. With Remake the story the clip is only read; it is never given to the video model, so the faces come from your own character sheets.
Can I edit the prompts it writes?
Yes. Every shot in the storyboard is text you can edit before you render, and you choose the visual style in Step 1.
What should a good video prompt include?
Framing, camera move, who is in frame, the action, the setting and light, the look, and any dialogue, one shot at a time.
Read a clip into a storyboard
Upload an MP4 or MOV, see the price of the read first, and get every shot written up.
Updated November 19, 2026