Consistent Character Prompt Generator

This consistent character prompt generator writes one image prompt per shot, and each one repeats your AI persona’s face, hair and look word for word, with the reference photo line on top. It is for AI influencer and AI character pages that need the same face in every post.

Free · No signup · Runs in your browser

1 · Identity
Woman or man
2 · Shots

    A shot can change the outfit (“in a wool coat”). The face, hair and signature detail stay the same in every prompt.

    Prompts

      How to use the Consistent Character Prompt Generator

      1. Fill in the identity, or press Random identity and change what you want.

      2. Write up to 20 shots, add ready-made ones, or fill six at random. Choose a photo style and aspect ratio.

      3. Press Generate prompts. Attach the persona’s photo, make prompt 1 and check the face, then go down the list.

      Why the identity is repeated word for word

      An image model makes each picture on its own. It only knows what that prompt and the attached images tell it. A face holds across a set when every prompt says the same thing about it.

      • The same words every time. OpenAI’s prompting guide for its image models suggests repeating the list of what to keep on every new image, so the person drifts less. The identity block here is that list.
      • The reference photo on every prompt. A text description alone gives a new face each time. The photo carries the face; the block backs it up.
      • Change the shot, not the person. Each prompt changes only the place, the action and the light. Clothes can change from shot to shot; the signature detail stays.
      • Start over from the block when a feature slips. Google’s guide for its image models suggests going back to a full, detailed description when a character drifts after many edits.

      How to keep the same face across a whole video

      A video is a run of short clips, and each clip starts from one image. Get the images right and the clips follow.

      1. Start from one clear photo. A sharp photo of the persona’s face. That photo is the face from now on.
      2. Make the first image from that photo. Attach it as the reference image with the first prompt.
      3. Make each next image from the photo plus the image before it. The photo holds the face. The last image holds the outfit, the room and the light.
      4. Redo any image where the face changed. Do it before you turn the image into video. An image-to-video model starts from that image, so a changed face carries into the clip.

      Make your AI persona from one reference image, and start every video from that face. How AI personas work →

      Consistent Character Prompt Generator questions

      How do I keep my AI character consistent in every video?

      Start from one clear photo of the persona. Make the first image of every video from that photo, and each next image from the photo plus the image before it. Redo any image where the face changes, before you turn it into video.

      Why does my AI character’s face keep changing?

      Because each image is made fresh. A text description alone gives a new face each time, and small changes creep in from image to image. Attach the same reference photo to every prompt, add your last good image next to it, and redo any image where the face changed.

      Do I need a reference photo?

      Yes, for the same face. The identity block keeps hair, build and look steady, but only an attached photo carries the exact face. Make one good photo first and attach it to every prompt.

      Make the first photo with the AI Persona Generator →

      How many reference images can I use?

      It depends on the model. Nano Banana 2 takes up to four images of people for character consistency, by Google’s docs. One clear face photo plus your last good image is usually enough.

      Can my persona wear different clothes in each shot?

      Yes. Write the outfit into the shot, such as “in a wool coat and scarf”. Keep one signature detail, like glasses or earrings, so the person still reads as the same at a glance.