Artvizon is credit-based, and the system is deliberately plain: you buy credits, they do not expire, and every generation shows its exact cost before you confirm. This article is about spending them well.
What actually drives the cost
Three choices, made at generation time.
The model. This is the largest factor by a distance. Video models are priced very differently by the providers who run them, and a premium model costs several times what a standard one costs for the same number of seconds. Choosing the premium model for a test you are going to throw away is the most common way to waste credits.
The duration. A multi-chapter video of 36 seconds is assembled from three 12-second chapters, so it costs roughly what three short clips cost. Length is not free.
The resolution. 720p by default, 1080p on the premium models.
Because these interact, we do not publish a table of costs per video: it would be true for one configuration and misleading for the rest. The exact figure for the generation you are about to run is on screen before you confirm it, and that figure is the authority.
For current pack sizes and plan prices, see the pricing page, which is always up to date in a way an article cannot be.
Spend the least on the decisions that matter least
The pattern that saves the most credits is simple: test cheap, publish well.
- Explore on a standard model, short. You are checking whether an angle works, not producing a final file. Three cheap tests teach you more than one expensive one.
- Choose the winner.
- Regenerate the winner at the quality and length you actually intend to publish.
Most people do this in the opposite order, generating their best guess at full quality first, and then paying again when it turns out the angle was wrong.
Get the input right before you spend anything
The cheapest quality improvement available costs zero credits.
- Use clean product images. The plan can only work with what it is given, and a soft or cluttered source constrains every generation you make from it.
- Say who the buyer is, in one sentence. It changes the whole plan.
- Fix a machine-translated description before generating. Two honest sentences beat a paragraph of keyword soup.
Fixing these after three failed generations is the expensive route to the same place.
Correct in the chat rather than starting over
If a result is close but wrong, say what is off: "too slow", "start with it in hand", "warmer light". The agent replans rather than beginning again, and you keep the context you have already established.
When a generation fails
Credits are returned automatically, and the agent tells you in plain language what happened. If a refund has not appeared after a few minutes, open a support ticket and it will be corrected by hand. A technical failure is our problem to absorb.
The one thing to remember
There is no unlimited plan, and there is no version of this where generating without thinking is free. The tool removes the production cost of a video, not the cost of the video. Deciding what to make is still the part that determines whether the spend was worth it.
Related reading: Pricing and A/B testing product videos properly.