Raul.

Case study 01 · AI Influencer

Sasha Kirana, a creator who doesn't exist.

Sasha is an AI-generated influencer I created to promote affiliate products. Every face, outfit and scene is generated — but the goal is content that feels genuinely human.

01

Video 1 — Maybelline lip tint

Product

Lip tint

Tools

ChatGPTLeonardo (Seedance)CapCut

A casual “get ready with me” beauty clip. The challenge: make a product demo look like it was filmed on a phone in a real bedroom.

Watch on Instagram

Step 01

Generate the first frame

I combine a character sheet and a product photo as references with a text prompt. The resulting image becomes the opening scene of the video — it locks in Sasha's face and the exact product.

Sasha Kirana character sheet
Character sheet
Lip tint product photo
Product photo

Step 02

Animate it with a detailed prompt

The generated image is used as the first scene, then a very specific prompt directs the motion, lighting, camera angle and the “no AI look” feel.

Prompt

“Close-up selfie video in a cozy bedroom at 10 AM, realistic Indonesian-Chinese woman Sasha Kirana sitting in front of a vanity mirror, wearing a white ribbed tank top and white short pants, natural messy long black hair, bare face no makeup, soft matte pink nude lips after using lipcream, holding a Maybelline SuperStay Matte Ink lipcream close to the camera with one hand. She slowly and naturally wiggles the lipcream product toward the phone camera while maintaining eye contact, subtle hand movement only, slight natural breathing, soft blinking, tiny head tilt, realistic human motion. Camera is placed on the vanity table with a slightly low side angle, handheld iPhone front-camera aesthetic, natural sunlight entering the bedroom, cozy warm tones, candid social media vibe, no AI look, realistic skin texture, slight camera micro-shake, natural hair movement, authentic TikTok beauty content style, shallow depth of field, smooth feminine motion, ultra realistic smartphone video quality, 9:16 vertical format.”

Step 03

Edit in CapCut

Once every scene is generated, I assemble them in CapCut, trim the pacing and add light edits to finish the video.

All generated lip tint scenes

Challenges

  • 01The lip colour after applying the tint didn't always match the real product shade.
  • 02Text and branding on the packaging get distorted — especially on small products.
Before and after lip tint

Before & after reference used to keep the lip colour consistent

02

Video 2 — Sportswear

Product

Sportswear

Tools

Figma WeaveKlingKling Motion ControlChatGPT

For this campaign I built a reusable workflow, so every new product video takes minutes instead of hours.

Kling Motion ControlInstagram

Step 01

Build the workflow once

In Figma Weave I connected every node — from product cut-out to final video — into one pipeline.

Full Figma Weave workflow

Step 02

Isolate the outfit

A screenshot from an e-commerce listing is regenerated into a clean outfit on a plain white background.

Outfit capture

Step 03

Choose a location

I add the background scene where Sasha will appear.

Location background

Step 04

Generate the photo

The workflow combines character, outfit and location into a single photo.

Generated photo

Step 05

Pick the product highlight

I decide which item the video should focus on.

Product highlight

Step 06

Turn it into video

The final node animates the photo into a short product video with Kling.

Generated video node

Step 07

Motion control with Kling

For motion control I feed a slowed-down reference video plus the location (red circle). The blue circle shows the result — Sasha copying the exact movement.

Kling motion control setup

Challenges

  • 01Kling V3 sometimes still produces footage that looks slightly AI-generated.
  • 02With motion control, the background can look separated from the character.
03

More examples

Product

Parfume — Evangeline Black Sakura

Product

Baggy Jeans

Product

Flare pants

Next case study

AI UGC Ad Strategy