HODLer: a meme character that never drifts
Model training, a playground and an AI agent for TON by Telegram
TON’s crypto media needed a steady stream of drawn content starring the HODLer meme character: the face and helmet identical in every image, while the outfit, place and situation change. We trained our own model, built a playground for the designers and an AI agent that illustrates Telegram posts by itself. Done in January–March 2025, when a character like this wouldn’t hold without a model of its own.
01
Face and helmet always, everything else new
The character comes down to two things: the meme’s recognizable face and the helmet. They have to match in every generation, or it isn’t HODLer anymore. The wardrobe, the setting and the situation from the post are new every time.
02
Eight training cycles
We hand-drew a starting dataset of 100+ character variations and trained a model on it. Each version went through 50+ test generations, and the best results fed the next cycle’s dataset. Over eight cycles the model learned to hold the face and helmet across any look, background and situation with under 5% error, and to mix styles without losing the character.
Dataset100+ hand-drawn character variations
Selectionthe best results feed the next cycle’s dataset03
A playground for the designers
Alongside the model we built a playground where the team’s designers tune and test generations in a simple interface before an image goes out.
04
An AI agent draws for every post
The last step automated posting: an AI agent reads a Telegram post and generates a HODLer image for it on its own. Content goes out with no manual moderation.
05
What has changed since
Today many tasks like this need no training: current models hold a character or a style from a few references. We train a custom model where references fall short: a strict character, a complex brand style or a high volume of content.
