Virtual persona Alina

A virtual influencer built on my own face and voice, kept consistent across photo, video and voice.

Scope

Visual identity · cloned bilingual voice (EN/ES) · documented content pipeline

The challenge Design and launch a virtual influencer as a complete product — not a single asset, but a consistent identity that holds up across image, voice and written content.
How it was made
The prompts
Every version, what failed and what fixed it → Prompt Lab
Volume
One scene, eight frames.

One base prompt fixes the scene, the light, the wardrobe and the hair. One line per frame changes only the angle and the expression — and the face holds across all eight.

{{ f.num }} · {{ f.short }}
Base prompt · all eight
The line that changes
{{ f.num }}
What failed

Frame 04 took three tries. Seen from behind, the blazer slipped off a shoulder, then turned into a sleeveless waistcoat. A wardrobe check naming both sleeves and full coverage fixed it; frames 05 to 08 carry a shorter version of the same line.

Cut for social

9:16 · 22 s · sound off · no new generations

The same study, cut as a 22-second vertical teaser: a hook, the method, what broke and the fix. Written to be read with the sound off, with the captions kept clear of the platform buttons.

What didn’t work

The character reference held Alina’s face but carried her default styling: early frames came out with her hair tied up and a draped coat that read as cosy rather than formal. Clothes stayed the weak point — the face holds from the reference; the wardrobe has to be written down, frame by frame.

The result

A consistent digital persona with a locked, reproducible visual identity, a cloned bilingual voice (EN/ES), and a documented production pipeline — with her own live website and social presence.

The learning The hard part of a virtual creator isn’t any single render — it’s consistency at scale. A documented pipeline is what turns scattered AI outputs into a believable, repeatable identity.
Next: Elena →
Download CV
Virtual persona Alina

A virtual influencer built on my own face and voice, kept consistent across photo, video and voice.

Scope

Visual identity · cloned bilingual voice (EN/ES) · documented content pipeline

The challenge Design and launch a virtual influencer as a complete product — not a single asset, but a consistent identity that holds up across image, voice and written content.
How it was made
The prompts
Every version, what failed and what fixed it → Prompt Lab
Volume One scene, eight frames.

One base prompt fixes the scene, the light, the wardrobe and the hair. One line per frame changes only the angle and the expression — and the face holds across all eight.

{{ f.num }} · {{ f.short }}
Base prompt · all eight
The line that changes
{{ f.num }}
What failed

Frame 04 took three tries. Seen from behind, the blazer slipped off a shoulder, then turned into a sleeveless waistcoat. A wardrobe check naming both sleeves and full coverage fixed it; frames 05 to 08 carry a shorter version of the same line.

Cut for social

9:16 · 22 s · sound off · no new generations

The same study, cut as a 22-second vertical teaser: a hook, the method, what broke and the fix — written to be read with the sound off.

What didn’t work

The character reference held Alina’s face but carried her default styling: early frames came out with her hair tied up and a draped coat that read as cosy rather than formal. Clothes stayed the weak point — the face holds from the reference; the wardrobe has to be written down, frame by frame.

The result

A consistent digital persona with a locked, reproducible visual identity, a cloned bilingual voice (EN/ES), and a documented production pipeline — with her own live website and social presence.

The learning The hard part of a virtual creator isn’t any single render — it’s consistency at scale. A documented pipeline is what turns scattered AI outputs into a believable, repeatable identity.
Next: Elena →