Illuminus7/Comfy_Int8_Fast_Workflows

ComfyUI int8-Fast workflows

Sample ComfyUI workflows for ComfyUI-AppleSilicon-int8-Fast, which runs int8 ConvRot models fast on Apple Silicon (M5). Install that node pack first; its README lists every model and LoRA these workflows load, with folders and download links.

Workflow	Model	LoRA
krea2_turbo_t2i.json	Krea 2 text-to-image	rgthree LoRA stack, empty by default (needs rgthree-comfy)
minimax_h3_t2v_mac.json	MiniMax H3 text-to-video	MiniMax H3 turbo LoRA
minimax_h3_i2v_mac.json	MiniMax H3 image-to-video	MiniMax H3 turbo LoRA
ltx2_5_t2v_mac.json	LTX 2.5 text-to-video	optional slot, set to None
ltx2_5_i2v_mac.json	LTX 2.5 image-to-video	optional slot, set to None

The LTX 2.5 LoRA slot uses the node pack's "Load LoRA (Bypass, Optional)" node: leave it on None, or pick a LoRA and set its strength on the main node.

The image-to-video workflows need a start image of your own in the Load Image node. Prompts are empty; write your own on the main node.

Tested with ComfyUI v0.40 and PyTorch 2.13 on an M5 Max (48 GB).