AnimeForge

Generating an image or a short clip is easy. A consistent finished piece is not. Research, writing, shot planning, continuity, assets, audio, subtitles, and revision all have to hold together. I built AnimeForge to treat that as engineering rather than a sketchbook of prompts.

The backend is FastAPI with async init, a database layer, provider registries, and separate APIs for research, ideas, projects, generation, and reports. The frontend is Next.js, React, and TypeScript. Providers stay loosely coupled so I can swap Ollama, OpenAI-style APIs, or ComfyUI-style workflows without redesigning the product around one vendor.

Long-running generation needs state, retries, monitoring, recovery, and reports. Continuity across characters and scenes is its own problem. That is the work I care about: useful software around models, not inventing a new model from scratch.