Describe it.
Play it. In one prompt.
Seele02
The First game-native harness agent
One workspace. One agent. Generate images, video, 3D assets, code, and complete games — end to end from a single prompt.
Powered by Seele02 Multimodal Foundation Model and Seele World Model

using UnityEngine;
public class BossController : MonoBehaviour {
[SerializeField] private float aggroRange = 12f;
private State _state = State.Idle;
private Transform _player;
void Update() {
var d = Vector3.Distance(transform.position, _player.position);
if (d < aggroRange && _state == State.Idle)
TransitionTo(State.Chase);
}
}A cloud workspace where you generate images, video, 3D assets, game builds, and web projects — all in one place. Files, preview, agent chat, and runtime in one tab. Already trusted by 1M+ creators. Free to start.
YOU
Laying out a 12×12km post-collapse industrial valley. Core is a refinery with tall silos for LOS cover; NE helipad for fast extract, SW rail yard for silent extract. Tier-3 vault under the silo cluster, two tier-2 rooms at the pump house and control tower. Sandstorm weather volume, visibility 80m…
YOU

From prompt to payout — SeeleAgent generates images, 3D assets, video, gameplay code, marketing materials, and complete games. Every card below shows one piece of that pipeline.

LYRA › Welcome, traveler. What brings you to the Amber Coast?
YOU › I’m looking for the lost shrine.
LYRA › The shrine lies beyond the▌



AI-generated videos, posters, and social posts — ready to ship.




| PARAM | OLD | NEW | |
|---|---|---|---|
| boss.hp | 800 | → | 720 |
| drop_rate | 12% | → | 18% |
| enemy.spd | 3.2 | → | 2.8 |
| world.era | 4th | → | 5th |
✓meta + structured data
✓12 backlinks submitted
→X · TikTok · 小红书

In-house multimodal foundation models, world models, and Agent systems
A sustainable technical backbone for the next generation of game production
Native multimodal foundation model with a Mixture-of-Transformers (MoT) architecture — unified representation of text, 3D, and space with native understanding, generation, and multi-turn editing. Tool-call ready as a general Agent base.
Treats mesh as a native language for multimodal models — understand objects, generate geometry, and keep editing across long context without losing identity.
World model with long-term 3D context memory — predicts how game worlds evolve across time with spatial consistency. The foundation for truly persistent, living games.
A unified 3D-native multimodal model — mesh understanding, generation, and context-aware multi-turn editing inside a single Mixture-of-Transformers architecture.
READ POST →A hybrid world model that pairs explicit 3D memory with 2D video generation — preserving geometry, sustaining long-term coherence, and enabling controllable scene evolution.
READ POST →In this post we introduce NeuralG-Bridge — a new world-model training paradigm that bridges game engines and video generation, using ground-truth engine state to supervise generative models.
READ POST →Instant access to Seele02-Pro, faster generation across all modalities, and full ownership of everything you create.
How SeeleAgent builds games, images, videos, 3D assets, and more.
Hire SeeleAgent. Open the workspace. Ship your first game, video, or scene — tonight.