Z-Image: Text to Image

# Python 3.10+: pip install comfy-sdk==0.4.0
import os

from comfy_sdk import Comfy

api_key = os.environ.get("COMFY_API_KEY", "").strip()
if not api_key: raise RuntimeError("Set COMFY_API_KEY")
os.environ.setdefault("COMFY_BASE_URL", "https://cloud.comfy.org")
client = Comfy(api_key=api_key)
workflow = client.workflows.from_json({
    "9": {
        "class_type": "SaveImage",
        "inputs": {
            "filename_prefix": "z-image",
            "images": [
                "76:65",
                0
            ]
        },
        "_meta": {
            "title": "Save Image"
        }
    },
    "76:62": {
        "class_type": "CLIPLoader",
        "inputs": {
            "clip_name": "qwen_3_4b.safetensors",
            "device": "default",
            "type": "lumina2"
        },
        "_meta": {
            "title": "Load CLIP"
        }
    },
    "76:63": {
        "class_type": "VAELoader",
        "inputs": {
            "vae_name": "ae.safetensors"
        },
        "_meta": {
            "title": "Load VAE"
        }
    },
    "76:65": {
        "class_type": "VAEDecode",
        "inputs": {
            "samples": [
                "76:69",
                0
            ],
            "vae": [
                "76:63",
                0
            ]
        },
        "_meta": {
            "title": "VAE Decode"
        }
    },
    "76:66": {
        "class_type": "UNETLoader",
        "inputs": {
            "unet_name": "z_image_bf16.safetensors",
            "weight_dtype": "default"
        },
        "_meta": {
            "title": "Load Diffusion Model"
        }
    },
    "76:67": {
        "class_type": "CLIPTextEncode",
        "inputs": {
            "clip": [
                "76:62",
                0
            ],
            "text": "A dramatic 1980s avant-garde fashion editorial portrait captures a young adult with an arresting, graphic presence, featuring an ultra-sharp bleached geometric buzzcut, stark black angular eye paint, and an exaggerated, structured metallic-gold lamé power suit with sculpted pagoda shoulders. The figure stands off-center across an endless expanse of cracked obsidian salt flats that stretch toward the distant, haze-shrouded horizon line. Towering, razor-sharp crystalline spires rise from the dark mineral crust under a bruised violet twilight sky. Low-angle, harsh golden spotlighting cuts violently through rising sulfur steam, carving deep chiaroscuro shadows across the gleaming landscape and casting long, razor-thin silhouettes across the vast, uninterrupted ground. The atmosphere is tense, theatrical, and darkly sublime."
        },
        "_meta": {
            "title": "CLIP Text Encode (Positive Prompt)"
        }
    },
    "76:68": {
        "class_type": "EmptySD3LatentImage",
        "inputs": {
            "batch_size": 1,
            "height": 768,
            "width": 1344
        },
        "_meta": {
            "title": "EmptySD3LatentImage"
        }
    },
    "76:69": {
        "class_type": "KSampler",
        "inputs": {
            "cfg": 4,
            "denoise": 1,
            "latent_image": [
                "76:68",
                0
            ],
            "model": [
                "76:70",
                0
            ],
            "negative": [
                "76:71",
                0
            ],
            "positive": [
                "76:67",
                0
            ],
            "sampler_name": "res_multistep",
            "scheduler": "simple",
            "seed": 770044821593082,
            "steps": 25
        },
        "_meta": {
            "title": "KSampler"
        }
    },
    "76:70": {
        "class_type": "ModelSamplingAuraFlow",
        "inputs": {
            "model": [
                "76:66",
                0
            ],
            "sampling": "flow",
            "shift": 3
        },
        "_meta": {
            "title": "ModelSamplingAuraFlow"
        }
    },
    "76:71": {
        "class_type": "CLIPTextEncode",
        "inputs": {
            "clip": [
                "76:62",
                0
            ],
            "text": ""
        },
        "_meta": {
            "title": "CLIP Text Encode (Negative Prompt)"
        }
    }
})

# run() also passes api_key to the partner nodes in this workflow.
job = client.run(workflow, api_key=api_key)
for output in job.get_outputs("9"):
    output.to_file(output.name)