Tutorial · Video

How to make the best AI videos

The difference between a forgettable clip and a cinematic film comes down to four choices. Here they are.

1. Start from a strong image

The best AI videos are image-to-video. Upload a great reference photo or generate a cover image first, then use it as the opening frame. A clear, well-composed start frame gives the video direction.

💡 How in 7stories

In the studio, upload a photo (step 6) and pick + Video or Both (step 7). The first photo becomes the video's opening frame.

2. Describe motion, not just the scene

Tell the video engine what moves and how. "The camera slowly pushes in… the flag waves… the couple turns toward the light." Motion words (pan, zoom, drift, fade, turn) matter more than adjectives.

3. Pick the video engine

  • Sora 2 (default) — strong, expressive motion.
  • Veo 3 — premium cinematic realism.
  • Seedance 2.5 — advanced control and quality.
  • Kling 1.6 — great motion and character detail.

Want the same person across multiple scenes? That's the consistent-characters workflow.

4. Choose resolution

In step 13, pick 720p for speed (cheaper, good for drafts), 1080p for final social posts, or 4K for large screens and advertising. Higher resolution costs more credits.

5. Regenerate until it moves right

Video is stochastic — the first take isn't always perfect. Use 🎲 Regenerate to get fresh takes on the same prompt until one matches your vision.

ℹ️ Related

See AI models for engine details, or the prompt guide.