You can generate a photorealistic human, and then animate her, with two text prompts. No camera, no model, no shoot. I asked Claude to write the prompts, pasted them into Leonardo AI, and got the image and the animation you see in this post.

Three steps.

One. Ask Claude (or any capable AI assistant) to write a detailed image prompt. The trick is specificity: age, skin, hair physics, lighting, camera language. Vague prompts produce generic AI faces.

Two. Paste the prompt into Leonardo AI and generate the image. Pick the best output.

Three. Take the same prompt into Leonardo’s video generation, add a negative prompt to kill the plastic AI look, and animate the image.

Same description, so the animation stays consistent with the image.

Negative prompt:

Four things carry the result.

Specific physical detail. “A few loose individual strands falling softly across her face” gives the model something concrete to render. “Beautiful hair” gives it nothing.

Motion baked into the description. “Hair moving gently as if caught in a light ocean breeze” matters twice: it adds realism to the still image and gives the video model a motion cue.

Camera and lighting language. Golden hour, shallow depth of field, editorial photography style. The model has seen millions of photos tagged this way, and it reproduces the look.

The negative prompt. Video models drift toward stiff hair, airbrushed skin, and studio flatness. Naming those failure modes keeps them out.

Change the subject, keep the structure, and you can produce consistent human characters for ads, social posts, or product mockups without a photoshoot.