Playbooks
Video · OpenAI

Sora

Text-to-video: short shots, described the way a film crew would.

Best for

  • short atmospheric shots
  • product motion
  • storyboard tests

Signature move

Write the shot list first and generate the hardest shot last.

How it wants to be spoken to

01

One shot per generation. Not a scene — a shot.

02

Describe subject, one action, camera move, light, mood, duration.

03

Keep the motion simple. One action per clip.

04

Reuse a reference image when you need the same thing twice.

Why it glitches

None of this is you doing it wrong. These are the shapes it reliably falls into — and the fix beside each.

Complex action breaks → split it into separate shots.

Continuity across clips is unreliable → lock a reference image.

Hands and text struggle → frame them out.

Camera moves get exaggerated → say “slow” and “subtle”.

A prompt that works there

Paste this into Sora
One shot, 6 seconds: [subject] [single simple action] in [setting]. Camera: a slow push in from [framing] to [framing]. Light: [warm late afternoon]. Mood: [calm]. No cuts, no text, no extra people.

Bring Tay your actual task and she'll shape it for Sora.

Practice with Tay