WEEKDAI 2 – Creating AI Characters with Generative AI

Mar 8, 2026

WEEKDAI Livestream No.2 – TOMLIN’S AI CHARACTERS, PART 1

In the second episode of WEEKDAI, we explore how we create AI-generated characters and turn them into structured assets that can be reused across different audiovisual projects.

WEEKDAI is our weekly live stream where we share experiments, workflows, and tools related to generative AI, creative technology, and next-generation audiovisual production.

Watch the full episode here.

Meet Two of Our TOMLIN CHARACTERS

In this stream we introduce two characters from our growing collection of TOMLIN CHARACTERS:

Maja – a Stockholm-based strategic thinker and underground DJ.
Tomasz – a charismatic late-50s creative with roots in skate and street culture.

These characters are part of our ongoing exploration of how AI-generated characters can become reusable visual assets for storytelling, visual prototypes, and future audiovisual projects.

 

From Character Image to Character Sheet

During the stream we demonstrate a simple character workflow:

  • Creating a base character using Recraft V4

  • Generating a character sheet with Nano Banana 2

  • Comparing how different AI models interpret the same character

  • Testing character consistency across multiple outputs

We also briefly show how we structure these experiments using ComfyUI pipelines, allowing us to iterate faster and experiment with different models and parameters.

 

Why We Do This

One of the biggest challenges with generative AI is character consistency.

By experimenting with structured workflows and model combinations, we explore how AI-generated characters can become repeatable and scalable assets, rather than one-off images.

This type of thinking also connects to our broader work with structured generative AI pipelines.

How to join

WEEKDAI is streamed live over Zoom, but access happens through our Discord community.

In Discord, we:

  • Announce upcoming live sessions

  • Share the Zoom link and a special passcode

  • Collect questions and requests before each stream

  • Chat under and after the stream

If you want to watch, participate, or influence what we explore, you need to be there.

When it happens

WEEKDAI is our weekly live stream where we explore generative AI tools, workflows, and creative experiments.

We stream every Friday at 12:30 (CET) for about 30–35 minutes, sharing real workflows, ideas, and experiments from inside the studio.

Learn more about WEEKDAI.

 

Open, live, and a bit messy

Everything we generate during WEEKDAI is built inside of ComfyUI using both Closed source and Open source tools and models. The sessions are live, unscripted, and sometimes unpredictable, and that’s exactly the point.

If you’re curious about generative AI, ComfyUI, or how GenAI workflows and pipelines work in real creative workflows, we’d love to have you join.

See you there live!

0 Comments

Submit a Comment

Your email address will not be published. Required fields are marked *

WEEKDAI 4 → Parametric Character Creation in ComfyUI

From prompts → systems
We generate characters using structured parameters and test outputs across models:
• Nano Banana 2
• Recraft V4
Same input → very different results
Watch:

#GenerativeAI #AI

People are missing this:

ComfyUI is the only tool in class positioned to ride the agent wave. Others are simply too closed off and unextensible. In a future where agents are integral to work, the artists who stuck with ComfyUI will be miles ahead of those who went elsewhere

WEEKDAI #3 → from empty room → styled interior

We:
• Define style
• Add kitchen layout
• Furnish + decorate

Not prompting.
Building systems.

🎥 Full stream:
https://www.youtube.com/watch?v=Ruvdgy6OCEg

@generaitr #GenAI #GenerativeAI #AISystem

WEEKDAI #3 → parametric prompting → full interior generation

No manual prompts.
Just structured inputs:
• Room type
• Style
• Time of day
→ system builds the scene
Then we decorate it + add characters.

🎥 Full stream:
https://www.youtube.com/watch?v=Ruvdgy6OCEg

#GenAI #GenerativeAI

Load More