Q&A · Getting Started
Quick answer
A good AI video prompt reads like a shot description: subject, action, setting, lighting, style, and camera movement, in concrete visual language. Compare "a dog" with "golden retriever puppy bounding through shallow surf at golden hour, slow-motion tracking shot, water droplets backlit" — the second gives the model a scene to execute rather than a guess to make. One clear scene per clip beats cramming several ideas into one prompt.
Camera vocabulary is the most underused lever: terms like "slow push-in," "aerial drone shot," "handheld close-up," and "tracking shot" map directly to motion the models learned from real footage.
Lighting and mood words (golden hour, overcast, neon-lit, soft window light) do disproportionate work, because they constrain the entire color and atmosphere of the output at once.
Iterate one variable at a time — change the camera move or the lighting, not both — and study working prompts: VdoBloom’s template gallery publishes full prompt text for every template.
VdoBloom starts you with 10 free credits — enough to put this into practice with no card required.
Open Text to Video tool