Digital Being — Roadmap

In the previous post, we stepped back from the idea of growing a humanoid pixel organism from a single cell. While the vision remains compelling, generating and evolving a capsule-based character from nothing introduces a number of high-risk assumptions. Rather than tackling all of that at once, we’re shifting to an iterative approach—building what we can confidently execute today, and deferring growth simulation to future iterations.

Roadmap

• Type 1 — SpriteDX (Completed)
  Fully animated, controllable vessels generated through a one-click pipeline. SpriteDX produces not just a reference character, but a complete set of animations (idle, run, jump, greet, etc.), making the character immediately usable in an interactive setting. However, these characters have no agency—they simply respond to external commands.

• Type 2 — Agent (Current Focus)
  We introduce agency. Characters begin to act on their own—moving autonomously, reacting to context, and eventually speaking through LLM integration. This phase also expands the animation state space to support richer behavior. A key assumption here is that a single reference image acts as the character's genome. Rather than explicitly modeling attributes like limb proportions or personality traits, we assume these are implicitly encoded in the image itself. The system interprets and expresses those traits through behavior and animation.

• Type 3 - Some of Parts (Future)
  Instead of treating the character as a monolithic entity, we decompose it into parts—head, hands, torso, etc. This enables more dynamic interaction and a broader range of motion. However, it introduces significant challenges: layering clothing, modeling soft features like cheeks, generating consistent rigged animations, and resolving rig mismatches. We expect to revisit this once Type 2 provides a stronger foundation.

• Type 4 - Growth (Future)
  The original vision: growing a complete organism from a single cell using local rules. This remains a long-term goal, to be explored once the earlier stages are stable and better understood.

— Sprited Dev 🐛