No still, no source clip. This page starts with a sentence and puts five video models behind one prompt box: Veo 3 at 300 credits, Seedance at 500, Wan 2.5 at 320 to 960 by resolution and its Fast variant at 460 or 690, and Sora 2 at 60 for ten seconds or 90 for fifteen. The credit cost for each run is on screen before you generate.
Use this route when the idea exists as words, not as a photograph or a finished cut. There is no upload field here, so nothing anchors the composition except the prompt and the parameters you choose. That makes it the fastest of the three video routes to test: you can try a scene, reject it and rewrite one line without handling media. The sibling routes solve a different problem. Image to video needs a frame before it can start, and video to video needs footage to restyle or extend. If you already have either, use those pages. If you have a sentence, stay here.
The five models are not interchangeable, and the credit price is the simplest way to choose between them. Veo 3 costs 300 credits and returns an eight-second clip at 720p on the Fast model or 1080p on Quality, in 16:9, 9:16 or Auto. Seedance costs 500 credits and runs from 3 to 12 seconds at 480p, 720p or 1080p, with a wider set of aspect ratios. Wan 2.5 costs 320, 640 or 960 credits per five seconds at 480p, 720p or 1080p and doubles for ten seconds; its Fast variant costs 460 at 720p or 690 at 1080p per five seconds, with no 480p option. Sora 2 costs 60 credits for a ten-second clip or 90 for fifteen on the standard tier, and runs landscape or portrait with a watermark-removal switch.
Pressing Generate opens the chosen model's own page, where the run is submitted. The credit cost and your balance are on the button before you leave this page, and a generation that fails to return a video is refunded at the platform level. That matters more here than on the other video routes. Prompt iteration is the whole point, so ten failed attempts on a 300-credit model would otherwise cost 3,000 credits and teach you nothing.
Read the panels before the price list. The five models do not expose the same switches: Veo 3 and Wan 2.5 let you choose a height, Seedance does not, and Sora 2 spends its controls on length and a watermark toggle instead. A model that hides a setting usually has a house treatment for it, so what you give up in control you get back in one fewer decision. For a preview that trade is an advantage. For footage that has to cut against material you already own, pick the model that lets you set the frame and the height yourself.
Only one model runs per submission. There is no switch that fires the same sentence at all five and lines the clips up, so a comparison is five deliberate presses, each of which leaves this page for the model you chose. Keep your sentence somewhere before you start, because the lab does not remember prompts already sent. Judge each return on movement first: whether the subject did the thing you named, in the direction you named, at the pace you implied. A clip that looks right and moves wrongly is a description problem, not a model problem.
Choose a model by budget
Veo 3 is 300 credits, Seedance 500, Wan 2.5 from 320 and Sora 2 from 60. The model selector shows the cost before you write anything.
Write one scene per prompt
Name the subject, the action and the camera move, then stop. Duration and aspect ratio are parameters below the prompt, so leave them out of the sentence.
Set duration and aspect ratio, then generate
Each model exposes its own controls: resolution where the model supports it, duration from 5 to 15 seconds, and the aspect ratios listed for that model. The button shows the credit cost and your balance.
Short social explainers
A 9:16 or 1:1 clip built around one clear action reads on a phone without sound. Veo 3 at 300 credits keeps the cost of a test low enough to run several versions.
Storyboard pitching
Generate one moving beat per scene so a client can feel the sequence before anyone books a shoot or builds a set.
Model comparison
Run the same sentence through Veo 3, Wan 2.5 and Sora 2 to see which model reads your direction the way you meant it.
Product and launch teasers
When no footage exists yet, a prompt-only clip can carry the first announcement while the real shoot is still being planned.
Model Selection
Describe what you want to generate
Cost: 300 credits | Balance: 0
Sample
Browse sample media