Gen-4.5 Text to Video: Widescreen 21:9 Clips from a Prompt

Last verified: October 19, 2026

Gen-4.5 turns a written shot description into a video clip. You can frame it in an ultra-wide 21:9 shape, or in square, portrait and landscape formats. The strength of Gen-4.5 text-to-video is how much direction one prompt can carry. The model excels at complex, sequenced instructions, so you can write detailed camera choreography, intricate scene composition, the exact timing of events and subtle atmospheric changes into a single prompt.

Working from text alone gives the model room to move. With no fixed opening frame, text-to-video handles complex motion sequences more effectively. That puts the whole shot in your words, so a prompt works best when it reads like direction for a camera operator, not a caption.

The form keeps the job focused: one prompt box with an Enhance Prompt helper for rough ideas, a Video Duration choice, an Aspect Ratio picker and a Seed field. Each run gives you back a clip to download. Pick the framing for where the clip will end up before you generate. Save the ultra-wide option for shots you've written with that much horizontal space in mind, such as wide landscapes, side-by-side staging or long tracking moves.

Explore more text to video

Gen-4.5 Inputs and Output at a Glance

What you can enter, choose and get back from each run.

Input

A text prompt of up to 2,048 characters

Video Duration

5 or 10 seconds

Aspect Ratio

1:1, 3:4, 4:3, 9:16, 16:9, 21:9

Output

1 video per run

Prompting Techniques for Better Gen-4.5 Clips

Describe Both the Picture and the Motion

Give every prompt two layers. The first is what we see: subject, setting, lighting and look. The second is how it moves: the subject's action, the environment and the camera. A dependable frame is '[Camera] shot of [subject] [action] in [environment]', followed by supporting detail. Anything you leave out becomes the model's creative choice. That's fine for background elements but risky for the main action, so spell that part out.

Time the Beats With Timestamps

When the order of events matters, tag each action with a time range, for example 0–3 seconds for the first move and 3–6 for the next. Keep a plain-language description of the whole shot alongside the timestamps; that combination gives the most control. Timestamps steer the general order and pacing but don't land to the exact frame, so use them for sequence and rhythm. Pick the longer video duration when the shot has several beats.

Start Lean, Then Refine One Detail at a Time

Start with only the most important visual and motion details, look at what comes back, then add detail where the clip falls short. Adjusting the prompt between runs is how Gen-4.5 results get refined. Change a single element per run, such as the light first and then the camera move, so you know which edit improved the take. When you're stuck on wording for the next layer, the Enhance Prompt button can suggest fuller versions to borrow from.

How to Make a Video From Text With Gen-4.5

Four steps take you from a written shot to a downloaded clip.

1

Write your shot in the "Prompt" box

Describe the scene in the "Prompt" box, which takes up to 2,048 characters. That's enough for a full shot description with subject, setting, light, action and camera move. If you're starting from a rough idea, click "Enhance Prompt" under the box to see fuller versions. Your text only changes when you pick a suggestion, so you can compare before committing.
2

Set "Video Duration" and "Aspect Ratio"

"Video Duration" offers 5 or 10 seconds. Use 5 for one clear action or a short plate, and 10 when a shot has several beats that need time to play out. "Aspect Ratio" offers 1:1, 3:4, 4:3, 9:16, 16:9 and 21:9. Choose 21:9 for a letterboxed, cinema-wide frame, 16:9 for standard video timelines, 9:16 for vertical social feeds, and 1:1 or 3:4 for feed posts. Then write your composition to fit that shape. The "Seed" field takes a number. Keeping it the same while you change one detail of the prompt makes before-and-after comparisons fairer.
3

Click "Generate"

Each run returns 1 video. Check the duration and frame shape once more before clicking "Generate", because getting either wrong means running the shot again.
4

Download the result

Watch the clip to the end before saving it. Look for props that vanish behind other objects or actions that happen out of order, then download it. If a take shows one of those slips, the troubleshooting fixes below explain what to rewrite before the next run.

Gen-4.5 vs Kling 2.1 Master: Which to Pick

Where the two text-to-video models on Vidofy differ, and where they're the same.

Frame shapes Six frame shapes are available: square, portrait, landscape, classic TV-shaped frames in either orientation, and an ultra-wide cinema frame. Kling 2.1 Master offers three frame shapes: square, vertical and standard widescreen. Pick Gen-4.5 when the edit needs an ultra-wide or classic TV-shaped frame; for square, vertical or standard widescreen, both offer the shape, so decide on how each handles your prompt.
Following detailed direction The model is documented as carrying out complex, sequenced instructions in a single prompt, including camera choreography, scene composition and the timing of events. Kuaishou describes the Master edition of Kling 2.1 as delivering superior motion performance and enhanced semantic responsiveness. Both makers stress how well their model responds to prompts, so run the same multi-step prompt through each and keep the one that holds your camera moves and action order.
Prompt, clip length and output Each run takes one text prompt and returns a single clip at a short or long duration. Kling 2.1 Master also takes one text prompt with an identical character limit and returns a single clip at the same two durations. These limits are identical, so they shouldn't drive the choice; decide on frame shape and on how each model handles your direction.
Use Gen-4.5 when the clip needs an ultra-wide or classic TV-shaped frame. For square, vertical or standard widescreen, both models share the same limits, so test one prompt on each and keep the take that follows your direction best.

Common Gen-4.5 Problems and How to Fix Them

Known ways prompt-only clips go wrong, and what to change before the next run.

An object vanishes after something passes in front of it

Gen-4.5 can lose track of objects: items may disappear or appear unexpectedly across frames, such as a cup vanishing after it has been briefly hidden. Rewrite the shot so the prop that matters stays in clear view. Have the passing person or object move behind it instead of in front, or have the camera keep it framed, then generate again. If the prop isn't essential, simplifying the scene around it is usually quicker than chasing a perfect take.

The door opens before anyone touches the handle

Effects sometimes arrive before their causes in Gen-4.5 clips, for example a door opening before the handle is pressed. Spell out the order in plain sentences with the cause first ('her hand presses the handle, then the door swings open'), give the cause a moment of its own, and run it again. The timestamp technique above gives you firmer control over which action comes first.

Every attempt succeeds, even when I want the character to miss

Gen-4.5 tends to let actions succeed more often than they should, so even a poorly aimed kick can still score a goal. Don't describe the attempt and hope for a miss. Describe the failure as the visible event, for example 'the ball sails wide of the left post and bounces off the advertising board', then generate a new take.

I wrote 'no camera movement' and the camera moved anyway

Negative wording works against you. Describe what you want to see, not what you want to avoid, because positive instructions get better results. Replace 'no camera movement' with 'locked-off static camera on a tripod' and 'no people' with 'an empty street'. Then run the prompt again with only that wording changed, so you can tell whether it fixed the problem.

My actions feel rushed, or one of them never happens

Each clip has a fixed length, and every action needs the time it would take in real life. Walking across a room needs about three seconds, not half a second. Count the actions in your prompt against the clip length you picked. Either cut secondary actions so the main ones have room, or switch to the longer video duration, then generate again.

Frequently Asked Questions

Can Gen-4.5 make a video from text alone, without a starting image?

Yes. The prompt is the only input here, and text-to-video creates footage directly, without making an image first. Because there's no starting frame to work from, every visual choice you care about has to be in the prompt.

Can I make vertical videos for Reels, Shorts or TikTok with Gen-4.5?

Yes. The Aspect Ratio setting includes a vertical shape alongside square, landscape and ultra-wide options, and you choose it before generating. Write the prompt for a tall frame: keep the subject in the centre and have the action move toward or away from the camera, so the key motion stays inside the narrow width.

What makes one Gen-4.5 generation cost more than another?

The credit cost of each run is calculated from the options you choose. Once you're signed in, it's shown above the Generate button. Video Duration is the setting that changes it, so check the figure after switching clip length. Suggestions from the Enhance Prompt button don't use credits.

Can I make a video longer than one clip with Gen-4.5?

Each run produces one clip at one of the lengths under Video Duration, so you build a longer piece as a series of shots. Write a prompt for each shot, generate them separately and cut them together in your editor. Text-to-video suits footage where exact character or scene consistency isn't the priority, so a recurring person can look different from shot to shot. Keep the same frame shape and the same wording for recurring subjects. Montages and establishing shots work best, because small differences there don't break continuity.