AI video models, compared on your own prompt

Why the model matters more than the prompt

A video generator is only as good as the model underneath it, and the models differ far more than their marketing suggests. A lightweight diffusion model will give you plausible motion in under a minute, but fine detail such as fabric, text on signs or reflections tends to smear. A full-size audio-video model renders those details convincingly and even produces a matching soundtrack, but it takes several minutes and costs many times more per second.

We expose that trade-off instead of hiding it. Every render on this site is tagged with the tier that produced it, every tier has a published credit rate per second, and the studio lets you queue two tiers on the same prompt so you can judge the gap yourself rather than taking our word for it.

The three tiers at a glance

Draft is the fast lane. It accepts clips from one to twenty seconds, outputs 720p or 1080p at 24 or 48 frames per second, and charges one credit per second at 720p. Use it to check that the camera move, subject and timing you described actually read on screen.

Standard is built on a compact audio-video model. Clips run four to fifteen seconds at 480p or 720p, sound is generated together with the picture, and the rate is six credits per second at 720p. For most social and explainer work this is the tier that ships.

Cinematic is the full-size model. It adds a 1080p option, noticeably steadier motion, more faithful lighting and the most coherent generated sound. At 720p it is twelve credits per second; at 1080p it is thirty. Reserve it for the renders that will be seen large.

• Draft: 1–20 s, 720p/1080p, 24 or 48 fps, from 1 credit per second
• Standard: 4–15 s, 480p/720p, generated sound, from 3 credits per second
• Cinematic: 4–15 s, up to 1080p, generated sound, from 6 credits per second

How to run a fair comparison

Keep everything but the tier constant. Same prompt, same aspect ratio, same clip length, same resolution where both tiers support it. The compare mode in the studio enforces this for you: you write the prompt once, pick a tier for side A and a tier for side B, and both renders are queued together under one comparison id.

Then look at specific things rather than an overall impression. Does the subject keep its shape across the whole clip? Do straight lines stay straight when the camera moves? Is the lighting direction consistent from the first frame to the last? Those three checks separate the tiers faster than any spec sheet.

Finally, decide per project rather than per account. A storyboard pass is a Draft job. The approved shot might be a Standard job. The one clip going in front of a client is where Cinematic pays for itself.

What the credit rates actually buy

A five-second Draft at 720p is five credits, which means the two free credits on a new account cover a two-second Draft test on your first idea. The same five seconds on Standard at 720p is thirty credits, and on Cinematic at 1080p it is one hundred and fifty. Running Draft first and only promoting a prompt that works is the single biggest saving you can make.

If a Standard or Cinematic render fails for a reason that is not yours, such as a model outage or a content filter on the provider side, the credits come back automatically. Draft renders are cheap enough that they are charged once started.