The look of the film
Choose one and reuse it in every shot. Consistency between clips is what makes a sequence read as one film rather than a pile of footage. If you don't know yet, leave it on Open — the model will make coherent choices for you.
Set this in Flow's own aspect control too. It appears in the prompt as a reminder, but the app setting is what actually crops the render.
Override the look
Only touch these if you disagree with the preset. Each is a real opinion about the image — leaving one open is a legitimate answer.
Realism cues stops it looking synthetic
The phrases that push output away from the smooth, plastic, too-clean surface. Four to seven works best — all sixteen dilutes each one. Hover any chip for what it fixes.
Framing & camera
Camera, in detail
Everything here defaults to open. Set only what you have a view on.
How you're generating this
Changes what gets written. Image-to-video and Extend deliberately drop most of the description, because the source frame already contains it.
Subject & action
Concrete nouns beat adjectives. Describe what a camera would actually record, not how it should make someone feel — the feeling comes from the framing and the light.
Performance and wardrobe
Light & air
If you set one thing on this page, set the light source. Naming a real lamp does more for believability than any adjective about mood.
Shaping the light
Weather and atmosphere
Sound
Veo generates audio natively. One line of dialogue, one effect, one background — crowded soundscapes come back as mush. Speech runs about 2.5 words a second, so an eight-second clip holds roughly sixteen words.
Beat sheet
Optional. One line per beat splits the clip into timed mini-shots inside a single generation, using Google's timestamp format. Three beats is the practical maximum for eight seconds — beyond that each one gets under two seconds, which isn't enough to establish anything.
Describe one continuous camera move across the beats, not three different ones. If the beats need two moves, they're two shots.
Characters
Google's documented method for keeping a person recognisable across shots: write one exhaustive description and paste the identical block into every shot they appear in, changing only the action and setting. This does that for you.
What the words mean
Written for someone with no filmmaking training. Nothing here is required reading — open a section when a control confuses you.