Mina LabsMina Labs Open the Studio

Consistent AI Character Video

The hard part of AI video is that the character changes between shots. Veo 3.1 Reference-to-Video takes up to four reference images of your subject and keeps them consistent across the clip, which is what makes a series of shots feel like one production rather than four unrelated ones.

Open the Studio → 20 free credits when you join · no subscription · failed generations refunded

What it costs

ModelCreditsPer
Veo 3.163 credits4s at 720p

Prices come from the same table the Generate button uses. New accounts start with 20 credits; failed generations are refunded automatically.

How it works

  1. Create your characterGenerate a character sheet or portraits in the Image Studio (Nano Banana Pro is great for this), or upload photos of an existing character.
  2. Open Video Studio and pick Veo 3.1 Reference to VideoUpload 1–4 reference images of your character.
  3. Describe the sceneWrite what your character does — the setting, the action, the camera move. Keep the character description consistent with your references.
  4. Generate more scenesReuse the same references for every new clip — the character stays consistent across your whole story.

Questions

How do I keep my AI character consistent across videos?

Upload 1–4 reference images of your character to Veo 3.1 Reference-to-Video on Mina Labs — it generates new scenes with the same face, outfit and style, without training a LoRA or fine-tuning anything.

Can I turn a character sheet into videos?

Yes — a character turnaround sheet works perfectly as a reference. Upload it, describe the scene, and the model keeps the character on-model in motion.

Which AI model is best for character consistency in video?

Google's Veo 3.1 Reference-to-Video is currently the strongest — it's built specifically to lock a subject from reference images. It runs on Mina Labs with pay-per-video credits and optional native audio.

Do I need to train a custom model?

No. Reference-to-video works instantly from your uploaded images — no dataset, no training time, no monthly fee.