Fizgig Adopts Driver-Style Model System; Qwen Image 2.1 Support Nearly Ready
shootthesound · reddit · 2026-09-27
Fizgig, an open-source local image model tool, is nearly done adding Qwen Image 2.1 support via a new driver-style architecture: each model gets its own driver holding model files, turbo LoRA settings and training adapters, so tools like Repair Studio and LoRA Explorer gain support automatically. Anyone will be able to add models (including older ones) via documented PRs. The author is weighing GitHub Discussions to avoid duplicate PRs for the same model and will publish a model-adding guide. They also trained a new Qwen Image 2.1 adapter on a large curated high-res dataset from their 20+ years of photography work, releasing Monday with noticeably sharper details.
More from coding & agent
- Adobe MAX: MCP across Creative Cloud, AI agents editing layers, Firefly Builder and Graph — umesh_ai · 2026-11-11
- Sydney Codex Community Buildathon set for Nov 14 with Cerebras-backed ultrafast compute track — paw_lean · 2026-11-11
- shadcn's improve hits 10k stars: let top models plan, cheap models execute code fixes — shadcn · 2026-11-11
- OpenAI open-sources Codex Security CLI and SDK for AI-driven vulnerability scanning and fixing — openai · 2026-11-11
- Claude Opus 5.5 runs an architecture search and finds AttnRes-Canon, a new top Transformer design — SonglinYang4 · 2026-11-11
- 5 research habits worth knowing: read papers backwards, reproduce first, wire coding agents via MCP — cneuralnetwork · 2026-11-11