Text to Image (New)

# 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({
    "60": {
        "class_type": "SaveImage",
        "inputs": {
            "filename_prefix": "z-image-turbo",
            "images": [
                "83:8",
                0
            ]
        },
        "_meta": {
            "title": "Save Image"
        }
    },
    "83:13": {
        "class_type": "EmptySD3LatentImage",
        "inputs": {
            "batch_size": 1,
            "height": 768,
            "width": 1344
        },
        "_meta": {
            "title": "EmptySD3LatentImage"
        }
    },
    "83:27": {
        "class_type": "CLIPTextEncode",
        "inputs": {
            "clip": [
                "83:30",
                0
            ],
            "text": "A surreal graphic poster set against a rain-drenched futuristic metropolis alleyway, where reflective puddles mirror flashing holographic billboards. In the center, a solitary bioluminescent lotus radiates an intense emerald and violet glow, its dew-covered petals pulsing with liquid light. Standing to the side, a slender young South Asian cybernetic courier with shaved temples and an iridescent teal topknot gazes at the blossom, wearing a soaked translucent vinyl trench coat over utilitarian dark mesh. Cascading rainwater and misty atmospheric vapor drift across the sprawling panoramic vista, blurring towering neon-lit architecture in the distance. Above the blooming aquatic flora, bold distressed typographic lettering reads \"AETHERIS\" in clean geometric stenciling. The visual combines wet urban grain, high-contrast shadows, and cinematic teal and magenta grading across the expansive composition."
        },
        "_meta": {
            "title": "CLIP Text Encode (Prompt)"
        }
    },
    "83:28": {
        "class_type": "UNETLoader",
        "inputs": {
            "unet_name": "z_image_turbo_bf16.safetensors",
            "weight_dtype": "default"
        },
        "_meta": {
            "title": "Load Diffusion Model"
        }
    },
    "83:29": {
        "class_type": "VAELoader",
        "inputs": {
            "vae_name": "ae.safetensors"
        },
        "_meta": {
            "title": "Load VAE"
        }
    },
    "83:3": {
        "class_type": "KSampler",
        "inputs": {
            "cfg": 1,
            "denoise": 1,
            "latent_image": [
                "83:13",
                0
            ],
            "model": [
                "83:28",
                0
            ],
            "negative": [
                "83:33",
                0
            ],
            "positive": [
                "83:27",
                0
            ],
            "sampler_name": "res_multistep",
            "scheduler": "simple",
            "seed": 528562900154240,
            "steps": 4
        },
        "_meta": {
            "title": "KSampler"
        }
    },
    "83:30": {
        "class_type": "CLIPLoader",
        "inputs": {
            "clip_name": "qwen_3_4b.safetensors",
            "device": "default",
            "type": "lumina2"
        },
        "_meta": {
            "title": "Load CLIP"
        }
    },
    "83:33": {
        "class_type": "ConditioningZeroOut",
        "inputs": {
            "conditioning": [
                "83:27",
                0
            ]
        },
        "_meta": {
            "title": "Conditioning Zero Out"
        }
    },
    "83:8": {
        "class_type": "VAEDecode",
        "inputs": {
            "samples": [
                "83:3",
                0
            ],
            "vae": [
                "83:29",
                0
            ]
        },
        "_meta": {
            "title": "VAE Decode"
        }
    }
})

# 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("60"):
    output.to_file(output.name)