Talk to the Doll (how to ai yourself)

From Image to a Sassy Talking Doll

Have you seen all those viral AI videos popping up in your feed lately, making everyone wonder, “How did they do that?”

Well… it’s actually way simpler than it looks.

Today I’m going to show you a fun, easy technique that anyone can pull off — no need to mess around with complicated settings.

We’re going to turn you (or any character you choose) into a cute doll version… and then bring it to life.

These kinds of videos are pure gold for social media — perfect for inside jokes with friends, and even for creative business use if you want to give your brand a unique twist.

To get the best result, we’ll be using Nano Banana 2 (our new image generation model) and Kling 3 to turn everything into a moving video.

View this post on Instagram

A post shared by The Electric Puma Ai Studio (@electric_pum_ai)

How to Create an AI Video: The Quick Guide

The process is made up of three simple steps. Pay attention 👇

1. Turn into a Doll

Take a good photo of yourself (or anyone you want).
The goal is to transform this image into a puppet (doll) version.

To do that, upload the image to Nano Banana 2 and use the following prompt:

Transform the provided reference image into a stylized 3D animated portrait in a plush, fuzzy, soft-texture aesthetic.

IMPORTANT RULES:
DO NOT copy colors from any predefined palette.
Extract ALL colors, lighting, and mood directly from the reference image.
Adapt the palette creatively (slight exaggeration or harmonization is allowed), but keep it faithful to the source.

STYLE TRANSFORMATION:
Convert the subject into a stylized, slightly androgynous animated character.
The character’s age MUST match the reference image exactly (do not generalize or shift age).

Simplify facial features:
• smooth skin
• minimal detail
• soft shading
• slightly exaggerated rounded nose
• half-closed or relaxed eyelids (deadpan / calm / mildly detached expression)
• slightly fuller cheeks and subtly enlarged eyes for a softer, more doll-like feel

TEXTURE & MATERIAL:
Hair must be based on the ORIGINAL hairstyle (length, shape, direction, and structure must match the reference).
Transform the existing hair into a voluminous, fluffy structure with a fuzzy, synthetic wool / plush-like texture.
Do NOT invent a new hairstyle.

Clothing should become soft, oversized, and fuzzy (like knitted/plush fabric).
Preserve original clothing colors/patterns, but reinterpret them in a playful, stylized way.

LIGHTING:
Use soft, diffuse studio lighting
Minimal shadows
Slight specular highlights on nose and hair
Clean, airbrushed finish

COMPOSITION:
Straight-on portrait
Tightly framed (head and upper torso)
Subject centered
Background must be flat and minimal, using a color derived from the reference palette

MOOD:
Deadpan, calm, slightly whimsical
Subtle emotional ambiguity

RENDER STYLE:
High-quality 3D animation look
Clean edges, soft gradients
Slightly toy-like / collectible figure aesthetic (subtle, not exaggerated)

OUTPUT REQUIREMENTS:
Preserve identity and pose from the reference image
Do not create a completely new character
Do not over-detail textures — keep them soft and stylized

2. Create Interaction

Now that we have the character, we want to create an image where both versions appear together — the real character and the puppet version.

To do this, simply copy the following prompt: Place both characters together

3. Add Motion

This is where the magic happens.

We take the image we created and move over to Kling 3.
There, we give it a prompt to turn the static image into a lively, funny video.

Cut 1 - Wide Medium Shot: Camera is static, eye-level. Both characters are visible. The real man stands slightly behind, looking at the doll. The doll faces forward. No camera movement — soft outdoor ambiance

CUT TO: Close-Up (Both Faces): Static close-up framing both faces. The real man slightly leans forward and asks:
"Which one of us looks more real?" — clean dialogue

CUT TO: Close-Up (AI Doll): Static close-up on the doll’s face only. The doll responds calmly:
"I do. You look like the cheap version of me." — flat confident tone

CUT TO: Medium Close-Up (Both): Static shot of both characters. The real man freezes, slightly confused and offended. No movement — short silence

CUT TO: Close-Up (Real Man): Static close-up on the real man. He slowly looks directly into the camera with a subtle, awkward expression. Hold for comedic timing — silence

Electric Puma Ai

I’m Lior — I use AI to turn ideas into something you can actually see and feel. I started out in written content, but today I’m drawn much more to the visual side of things. It’s a fast-moving space where new tools drop constantly and workflows evolve all the time — and that constant change is exactly what keeps me hooked. Most of what I share comes from hands-on experimentation — testing, failing, tweaking, and trying again until something clicks. Then I take those insights and break them down into simple, practical guides like this one. This isn’t about copy-paste. It’s about giving you a starting point — something you can build on, remix, and push further in your own way. It’s not a full AI course (not yet, at least), but it is meant to support anyone who sees AI not just as a tool — but as a creative playground. AI video is just the beginning. Let’s explore what’s possible together.