Skip to main content
Note: This content has moved. See the consolidated prompting guides in docs/guides/prompting/prompting-tips.mdx and the new deep-dive files: docs/guides/prompting/image-to-image.mdx and docs/guides/prompting/text-to-image.mdx. Effective prompts improve quality and consistency.

​
General guidance

  • Be specific and descriptive
  • Include movement, lighting, and style cues
  • Avoid contradictory instructions
  • Keep prompts under ~200 characters for best results

​
Examples

  • “A cinematic close-up of a hummingbird drinking nectar, golden hour lighting, shallow depth of field, smooth slow-motion”
  • “Futuristic city skyline at night with neon lights, gentle camera pan, light rain, cyberpunk style”
Info When using Image-to-Video, describe the intended motion relative to the source image, e.g., “subtle head turn and a friendly smile.”

​
Performance tips

  • Poll status every 5–10 seconds
  • Use exponential backoff for failures
  • Cache final video URLs for reuse
  • Prefer webhooks for production (see Webhooks guide)