Ovis Image Workflow

Ovis Image Workflow is a ComfyUI workflow for processing images with Ovis Image: it accepts a loaded image and Ovis model files, and includes steps for upscaling, face detailing, image comparison, and saving image results.

The graph includes image loading and resizing, upscaling, face detailing, image comparison, and image saving nodes.

Inputs include a loaded image plus ovis_image_bf16.safetensors, ovis_2.5.safetensors, and ae.safetensors.

Its image-processing steps include ImageUpscaleWithModel, FaceDetailer, Image Comparer, and Image Saver.

Place ovis_image_bf16.safetensors in ComfyUI/models/diffusion_models, ovis_2.5.safetensors in ComfyUI/models/text_encoders, and ae.safetensors in ComfyUI/models/vae.

Before use, update ComfyUI to the latest GitHub version and install the listed custom nodes.

Suggestion · not verified

The required model files are ovis_image_bf16.safetensors, ovis_2.5.safetensors, and ae.safetensors.

The listed optional model files are 4x_foolhardy_Remacri.pth for upscaling, bbox/face_yolov9c.pt for face detection, and sam_vit_b_01ec64.pth for segmentation.

Required custom nodes are ComfyUI-Manager, rgthree-comfy, ComfyUI-Easy-Use, ComfyUI-Image-Saver, ComfyUI-Impact-Pack, and ComfyUI-Impact-Subpack.

The listed settings are 20–50 steps with a default of 25, CFG around 5.0, sampler euler, and scheduler sgm_uniform or simple.

Listed sizes are 1024×1024 for 1:1, 896×1152 for 3:4, 832×1216 for 5:8, 768×1344 for 9:16, and 640×1536 for 9:21.

Before use, check that the required model filenames and folder placements match the listed setup.

Suggestion · not verified

Before enabling upscaling, face detection, or segmentation steps, check the placement of 4x_foolhardy_Remacri.pth, bbox/face_yolov9c.pt, and sam_vit_b_01ec64.pth.

Suggestion · not verified