Reference Pose Lora Workflow. Low VRAM GGUF.

This workflow is for creating images guided by a reference pose: it combines a reference image and text prompts with Flux.1 Kontext and the FluxKontext/refcontrol_pose.safetensors LoRA, then provides preview and saved image outputs.

The graph loads an image, preprocesses pose information, prepares reference latents, samples and decodes an image, and sends it to preview and save steps.

Inputs include a loaded reference image and text prompts; image sizing, cropping, and stitching are part of the processing.

Outputs include a preview image and a saved image.

Reference Pose Lora Workflow. Low VRAM GGUF is listed at https://civitai.com/models/1895984, published by nothingherern450454.

The listing describes it as designed for weak GPUs and reports use with an RTX 3060 12GB GPU and 16 GB DDR4 RAM.

Listed model files are ae.safetensors, clip_l.safetensors, flux1-kontext-dev-Q5_K_S.gguf, FluxKontext/refcontrol_pose.safetensors, and t5xxl_fp8_e4m3fn_scaled.safetensors.

Listed node packs are comfyui_controlnet_aux, comfyui-custom-scripts, ComfyUI-GGUF, rgthree-comfy, and teacache.

The main setup uses Flux.1 Kontext, flux1-kontext-dev-Q5_K_S.gguf, a GGUF UNet loader, and FluxKontext/refcontrol_pose.safetensors.

The graph includes FluxGuidance, Lora Loader Stack (rgthree), and TeaCache nodes, plus scaling, cropping, and stitching.

Before running it, verify that all five listed model files and all five listed node packs are installed and selectable.

Suggestion · not verified

Treat the reported RTX 3060 12GB GPU and 16 GB DDR4 RAM as hardware references, not guarantees for other systems.

Suggestion · not verified