ComfyUI-Batch-Render

ComfyUI Batch Render

Batch-render images across base × scenario combinations on ComfyUI.

Define a base (e.g. a character: checkpoint + prompt + LoRA(s)) and a list of scenarios (each its own checkpoint? + prompt + LoRA(s)), then render every combination automatically — no more manually swapping a LoRA, editing its prompt, and re-queuing one image at a time.

Base and scenario are the same shape (Layer = { enabled?, checkpoint?, prompt, loras[] }); disabled scenarios remain in the saved pipeline but are skipped when rendering. A render combines one base + one scenario. LoRAs are loaded by splicing a chain of stock LoraLoader nodes into your own exported workflow — no required custom nodes, no assumed folder layout.

Status

Early build. Implemented incrementally as composable tiers:

| Tier | What | Status | |------|------|--------| | 0 | Graph patcher (pure functions) | done | | 1 | Run engine + CLI | done | | 2 | ComfyUI plugin shell (web server) | done | | 3 | Web UI | done |

Design

Architecture and rationale live in the project notes (not committed). The short version:

• Bring your own workflow. Export your ComfyUI workflow in API format; the engine fills a few mapped slots (prompt, seed, model/clip splice point, checkpoint) and leaves the rest alone.
• Native & future-proof. LoRA loading uses stock LoraLoader nodes assembled in the workflow JSON; the only external contract is ComfyUI's stable /prompt + /ws API.
• Local & flexible. No assumptions about folder structure or ports; everything is read live or configured via a gitignored local config.yaml (see config.example.yaml).

Configuration

Copy config.example.yaml to config.yaml (gitignored) and set your ComfyUI host/port and paths.

License

MIT