TikTok AI Effects

From camera to social to conversation. Three waves of generative AI, designed to give creators new ways to tell their story.

TikTok's first wave of generative AI effects launched in 2022: AI Portrait in the camera, AI Profile in the social layer, AI-MOJI inside DMs. Each release expanded what AI could do for creators while sharpening the training pipeline behind it.

My Role

2D lead and visual direction. Set the bar each release had to clear, partnered with AI/ML engineering to refine the training pipeline behind it, and pushed what AI could do for creators with every wave.

Foundation — AI Portrait (2022)

The first wave. TikTok's earliest generative AI effects lived inside the camera. The model could only do image-to-image style transfer at the time, so the design problem was getting consistent quality out of a constrained capability, across multiple distinct styles. We set the creative bar each style had to clear, built training references the model could learn from, and shipped results that felt curated rather than generically AI.

Expansion — AI Profile (2022)

From camera to social identity. AI Profile extended the work out of the camera and into TikTok's social layer. Users uploaded a photo and got back a stylized profile picture, turning AI from a one-time effect into a piece of self-expression that lived on their account. The pipeline matured from AI Portrait, and the creative direction held the quality standard as the surface area grew.

Extension — AI-MOJI (2022–2023)

AI-MOJI brought generative expression into DMs. Personalized stickers and reactions generated from a user's own face, built around a single signature style. A deliberate constraint that made every output instantly recognizable, no matter who it came from.

Three challenges shaped the work: style consistency across every generation, rich expressiveness across the full emotional range, and real resemblance to the user's actual face. Solving all three at once is what the visual direction and training pipeline were built around.

Outcome

Each wave pushed the pipeline further: a constrained camera filter, then a social identity tool, then everyday conversational expression. That progress fed directly into the company's internal AI workflow.

Every feature went viral at launch, topping the effect ecosystem across global markets and driving a platform-wide wave of AI-generated content. The three waves proved out two core verticals for AI on TikTok, camera creation and social interaction, and showed that generative AI could ship at real platform quality.

Credits

2D Lead: DingDing Chung
AI Portrait — Design & Training: Sara Haas, Arian Tibbs, Jonathan Guzi, Yixin Zhao, Brandon Lai, Jae Bae, Michelle Catalanotto, Alex Noelke, Diana Lee
AI Profile — Visual Direction & Training: Jonathan Guzi
AI-MOJI — Visual Direction & Training: Jefrain Gallipoli, Jonathan Guzi