Glossary · Getting Started
Quick answer
A seed is the starting random number an AI model uses for a generation. The same prompt with the same seed on the same model produces the same (or very similar) output, while changing the seed produces a different variation. Seeds are how you reproduce a result you liked or explore alternatives without rewriting the prompt.
Generative models begin every image or video from structured noise, and the seed determines that noise pattern. Think of it as the dice roll behind the generation: lock the dice and you lock the variation.
In practice, seeds matter for two workflows. First, reproducibility — if a client approves a look, the seed lets you regenerate it. Second, controlled iteration — keep the seed fixed while tweaking one prompt word to see exactly what that word changes.
Not all hosted tools expose the seed parameter directly; when it is hidden, regenerating simply rolls a new random seed each time, which is why outputs differ run to run.
VdoBloom starts you with 10 free credits — enough to put this into practice with no card required.
Open AI Art Studio