Video generation

Generate AI video from text or from an image

The catalogue holds video models from several publishers — Kling, Seedance, Wan, MiniMax H3, LTX, Grok. Some start from text alone, some from a start frame, some from reference clips. They sit in the same canvas, on the same credit balance, and can be compared on the same brief.

What the models can do

Text to video, with no start frame

Several models — Kling, Seedance, Wan, MiniMax H3, LTX — generate a shot from the description alone. Useful when no visual of the subject exists yet.

From a still to a moving shot

This is the most common production route: you approve an image, then animate it. On the canvas the image generated at the previous step wires straight into the video node — no download, no re-upload.

References to hold a character or a set

Some models accept several reference images and reference clips, which lets you pin a character, a set and a camera move at once rather than hoping for them.

Sound along with the picture

Several models produce a soundtrack with the shot. For short social video that removes a round trip to a separate tool.

The formats platforms actually take

16:9, 9:16 and 1:1 depending on the model, with durations from 2 to 30 seconds and resolutions up to native 4K. The format is chosen before generating, not by cropping after.

Comparing costs less than being wrong

Video models do not render the same motion or the same material. Running one brief through two of them and choosing afterwards beats discovering the problem after the edit.

Online, with no edit bay and no graphics card

Generation runs in the browser and the compute happens server-side, at each model vendor. No render tying up a laptop overnight, no drivers to update, nothing to install. That is what makes AI video reachable from an ordinary office machine.

How to pick an AI video generator

Start with whether it accepts a start image: that is what separates a usable shot from a pretty loop you cannot cut into anything. Then look at the length and format actually reachable, not the ones on the home page. Then what comes after — a useful video goes through an upscale, an audio track, a vertical recrop. And finally whether you can try several models: none wins on every subject, and the wrong subscription costs more than comparing.

The whole chain

From the first frame to the delivered shot, in one space.

  1. 1 · Settle the start frameGenerate and approve an image with one of the catalogue image models. It sets the framing, the light and the material of the shot.
  2. 2 · Pick a video modelThe picker shows what each one takes as input, its durations, its ratios and whether it can generate sound. Some require an input image, some do not.
  3. 3 · AnimateThe image wires into the video node and the shot generates. The render comes back onto the canvas, next to the image it came from.
  4. 4 · DeriveThe same workflow reruns to produce the other formats or the other subjects in the series, without redoing the settings one by one.

Frequently asked questions

Can I generate AI video for free?
The free plan gives access to the whole catalogue, video models included, with a starting credit allowance. A video generation spends considerably more credits than an image, and the cost is shown before you launch it.
What is the maximum length of a shot?
It depends on the model: durations run from 2 to 30 seconds depending which one you pick. The picker shows the exact range for each. For a longer sequence you chain several generations inside a workflow.
Do I always need a start image?
No, but it depends on the model. Several accept text alone; others require an input image. The picker says which before you launch.
Do the models generate sound?
Some do, some do not. It is listed in each model’s capabilities, and it is an option on the generation rather than a separate step.
Can I produce vertical video for social?
Yes. 9:16 is available on most catalogue video models, alongside 16:9 and 1:1. The format is chosen before generating, which avoids a destructive crop afterwards.
Can I generate AI video without signing up?
No, an account is required: a second of generated video spends real compute, which has to be attached to someone. Signing up does not ask for a card — the free plan opens the whole catalogue with a starting credit allowance, enough to try several video models on your own shots.
Do I need a powerful machine or a graphics card?
No. Everything runs online, in the browser, with the compute happening server-side at each model vendor. An ordinary desktop is enough; rendering never ties up your machine.
Which free AI video generator is the best?
No model wins on every subject: the one that holds human motion best is not the one that renders a product or animated text best. The criteria that decide it are specific to video: accepting a start image to lock the frame, holding the stated duration without the subject drifting, producing a vertical format you can actually ship, and handling sound. The free plan is there to test those on your own shots before committing.

Generate a video

Start from text, or from an image you have just generated on the same surface.

Open the studioSee the modelsHow credits work

Keep reading