Text to Video

Script to Video: Turn a Written Scene into a Clip

Turn a written scene into a short video. Direct the subject, light and camera, compare generated versions, and use the chosen clip in your next story.

Explore the studio

SHOTTIDE / Text to Video01 — 03

01Write a focused prompt

02Choose the clip settings

03Review and assemble

Your direction. Every step.

THE WORKFLOW

From an idea to a result you can use.

Write a focused prompt

Describe what is visible, what moves and how the camera follows it. Choose landscape or portrait for your intended screen.

Choose the clip settings

The basic video editor uses Wan 2.2: select 480p or 720p and approximately 5 or 7.5 seconds. For LTX 2.5, open Model Board in the studio sidebar and choose a 5- or 10-second clip at 720p. Review the quoted credits before starting.

Review and assemble

Preview or download the result. In a storyboard project, choose a version for each shot and export them together with optional narration.

A STARTING POINT

Try this direction.

An editable prompt example, ready to make your own.

A red running shoe on wet asphalt at dawn. The camera slowly pushes in as a drop of water rolls across the fabric. Soft warm backlight, shallow depth of field, realistic material texture. No text.

Use this example

From a script line to a finished video clip

Mark the moment you want to see

Take this script line: “At closing time, Mara notices a light under the workshop door. She puts down her keys.” It contains enough material for two shots. For the first, keep the camera inside the dark hallway and let the line of light appear at the bottom of the frame. For the second, put the keys and Mara’s hand in view. Pick one of those beats for your first generation. Keep the original line in your notes so that, when you review the picture, you can tell whether the audience learned what the writer intended.

Write a first frame and a last frame

In the video workspace, describe a picture the model could start on. Try “A brass key rests between two fingers over a wooden counter, seen from the side.” Follow it with one change: “The fingers open and the key drops onto the wood; hold the camera still.” Decide how big the object should look at the end. If a cut needs to match the next scene, note the direction the hand leaves the frame. Save dialogue and voiceover in a separate line. That lets you adjust the reading without accidentally changing the visual instructions for the take.

Choose the take before spending credits

Read the action at the pace you want it to happen. Five seconds can be plenty for a key to fall, but a slow turn of a dial needs time for the audience to read its markings. Choose a supported duration and frame shape in the editor and review the credit quote. If you have a first-frame reference, check what is already visible in it. A close photo of a hand gives the model little information about the rest of the room. For an establishing view, begin with an image that actually includes the doorway and the surrounding space.

Review the cut, then export

Play the clip all the way through once. Then pause where the action changes: does the key touch the counter, or does it disappear behind the hand? Download a usable clip if that is the whole assignment. For a sequence, plan the shots in a project and select the take for each shot there. Review the last frame of one next to the first frame of the following shot. Add narration where it explains something the picture cannot show, then listen across each cut. Export the sequence and replay the file outside the editor to check the opening, the final frame and the sound.

Trailer: stop before the door opens

Script: “A night porter hears a bell in the empty hotel. The service lift climbs to his floor.” For a five-second teaser shot, frame the floor indicator and ask for the needle to move one mark and stop. A second shot can follow the porter’s hand reaching for the metal gate. Cut before the gate opens. In a trailer, the question is the payoff: what came upstairs? Check whether the needle stays readable at phone size. If it does not, move the opening viewpoint closer instead of adding an explanatory voice line that tells the viewer where to look.

Product demo: make one mechanism legible

Script: “Press the side catch and the measuring spoon folds flat.” Supply a reference showing the correct spoon and hinge. In the first view, leave space on the side where the handle will travel. In the next view, show the folded spoon beside a drawer so the buyer understands what the movement achieves. Allow about five seconds per view as an editing target. Inspect the joint before accepting a take: if the handle bends in the middle or travels through the bowl, it demonstrates a different mechanism. Keep the product claim tied to what the reference and the finished picture actually support.

Short story: carry a decision over the cut

Script: “A passenger sees a glove on the empty seat. As the train slows, she puts it in the lost property box.” Use the first clip for the discovery and the next for the return. Give the glove a detail that can survive a change in scale, such as one yellow stripe across the cuff. When reviewing the second clip, check whose hand holds it and whether the stripe is still visible. Two five-second settings give you roughly ten seconds of material before any changes to the edit. Watch the silent sequence first: someone who has never read the script should still understand the passenger’s decision.

Tutorial: leave the result on screen

Script: “Turn the paper template until the two corner marks meet, then pinch the fold.” Frame both marks before the hands enter. Separate the alignment and the fold into different takes if the fingertips cover the marks while moving. End the first view with the marks still visible so a learner can pause and compare their own paper. Read the instruction aloud and see if the action finishes before the sentence does. For a tutorial with exact positions, compare every step against the real object before publishing; a plausible hand motion can still produce the wrong fold.

Social short: design the first silent second

Script: “An envelope on a café table slides towards the edge. A hand catches it just in time.” Frame the edge and the envelope together for a vertical clip. Start close enough to see the overhang immediately, and leave space below for the arriving hand. One five-second take can work if the catch is clear. Watch the first second with the player muted, then cover the bottom of the screen with your thumb to approximate an interface overlay. If the catch disappears under it, change the framing. Record any spoken tag after choosing the take so the last word lands after the catch.

Separate subject and camera motion

“A cyclist rides past; the camera stays still” is more precise than several conflicting movement instructions.

Generate clips, then build the film

Each generation creates one short clip. Use a storyboard project to control the sequence, narration and final export.

Before you create

Can I try script to video free?+

You can draft a script and browse finished examples for free. Generating a video uses credits. Check your available balance and the quote for the selected clip before starting a take.

What should I give a script to video AI?+

Start with a subject, a visible action and the camera position. Give spoken lines their own space. For example, a sentence about a lost glove becomes a shot of a passenger picking it up; the voice can explain whose glove it is.

How do I make text to video from a script with several scenes?+

Split the script where the setting, viewpoint or action changes. Generate and review each scene separately. For a finished sequence, use a project to select the takes, arrange the shots and export the film.

What resolutions and durations are available?+

In the basic video editor, Wan 2.2 offers 480p or 720p with 81 or 121 frames at 16 fps: roughly 5 or 7.5 seconds. Model Board also offers LTX 2.5 at 720p with 5- or 10-second settings. These are individual clip settings; final film export offers 480p and 720p.

Does the generated video include sound?+

Wan 2.2 produces silent clips. LTX 2.5 in Model Board requests generated audio along with the video; listen to the result before using it. You can generate per-shot narration separately. Film export retains a clip’s existing audio when no replacement narration is selected; selected narration replaces that audio.

How can I compare the video models?+

Open Model Board and enable the Wan 2.2 / LTX 2.5 comparison. Both requests use the same prompt and optional first frame at 720p and approximately five seconds. This submits two paid tasks. Compare the motion, subject details and audio before choosing a take; model comparison does not guarantee either result will match your brief.

Can I revise a generated video?+

You can change the prompt and generate another version, then choose the version to use. A fine-grained editing timeline, motion brush and frame-by-frame editing are not included.

KEEP CREATING