Tutorial · Pro

Consistent characters across scenes

The #1 question in AI video: keep the same person looking identical in every shot. Here's the workflow — built into 7stories.

Generic text-to-video changes the subject between shots. The fix is a two-step pipeline: create a consistent character image first, then turn it into video. Two image engines are specifically great at character consistency: Nano Banana (Gemini 2.5 Flash) and Seedream 4.5.

The workflow

  1. Upload one clear photo of your subject in step 6. Face the camera, good light, full head and shoulders.
  2. Pick Nano Banana or Seedream as the image engine (step 8).
  3. Choose + Video or Both in step 7, and a video engine (step 9) — Veo, Kling, or Sora all work well.
  4. Describe the scene in step 5, keeping the same person's identity in the words ("the same young woman…").
  5. Generate. Because the reference photo anchors the subject, the image and the video's opening frame share the same face.

💡 Why this works

Nano Banana and Seedream are trained to hold a subject's identity when given a reference. Using that image as the video's start frame carries the identity into the film.

Tips for consistency

  • Use one clear, high-resolution reference photo — not a group shot.
  • Keep hair, clothing, and setting described the same way across scenes.
  • For a multi-scene reel, generate one image per scene from the same reference, then animate each.

ℹ️ Related

New to prompts? Read the prompt guide. Picking an engine? See AI models.