license	apache-2.0
language	en
tags	zen zen-lm image-generation image-editing diffusion
library_name	diffusers
pipeline_tag	text-to-image

Zen Artist

Diffusion-based image generation and editing by Zen LM

🤗 HuggingFace  |  📖 Docs  |  💻 GitHub

---

Introduction

Zen Artist is a 7B-parameter diffusion-based image foundation model from Zen LM by Hanzo AI. It delivers strong performance in both text-to-image generation and instruction-guided image editing, with particular strength in precise text rendering and multi-image composition.

Key Capabilities

• Text-to-Image Generation: Generate high-fidelity images from natural language prompts
• Instruction-Based Editing: Modify existing images using text instructions
• Text Rendering: Accurate in-image text, including multilingual characters
• Multi-Image Composition: Combine multiple reference images into a coherent scene
• ControlNet Support: Depth maps, edge maps, keypoint maps, and sketch conditions

Model Specifications

Attribute	Value
Parameters	7B
Architecture	MMDiT (diffusion-based)
Native Resolution	1024 x 1024
Maximum Resolution	2048 x 2048
Text Context	32K tokens
Model ID	zenlm/zen-artist
License	Apache 2.0

Quick Start

Install dependencies:

pip install transformers diffusers torch

Text-to-Image Generation

from diffusers import DiffusionPipeline
import torch

model_name = "zenlm/zen-artist"

torch_dtype = torch.bfloat16 if torch.cuda.is_available() else torch.float32
device = "cuda" if torch.cuda.is_available() else "cpu"

pipe = DiffusionPipeline.from_pretrained(model_name, torch_dtype=torch_dtype)
pipe = pipe.to(device)

prompt = "A cozy coffee shop at night, warm lighting, chalkboard sign with elegant lettering"
negative_prompt = " "

# Supported aspect ratios
aspect_ratios = {
    "1:1": (1328, 1328),
    "16:9": (1664, 928),
    "9:16": (928, 1664),
    "4:3": (1472, 1104),
    "3:2": (1584, 1056),
}

width, height = aspect_ratios["16:9"]

image = pipe(
    prompt=prompt,
    negative_prompt=negative_prompt,
    width=width,
    height=height,
    num_inference_steps=50,
    true_cfg_scale=4.0,
    generator=torch.Generator(device=device).manual_seed(42)
).images[0]

image.save("output.png")

Image Editing

import torch
from PIL import Image
from diffusers import ZenArtistEditPipeline

pipeline = ZenArtistEditPipeline.from_pretrained(
    "zenlm/zen-artist",
    torch_dtype=torch.bfloat16
)
pipeline.to("cuda")

image = Image.open("./input.png").convert("RGB")
prompt = "Change the background to a sunset beach scene"

inputs = {
    "image": image,
    "prompt": prompt,
    "generator": torch.manual_seed(42),
    "true_cfg_scale": 4.0,
    "negative_prompt": " ",
    "num_inference_steps": 50,
}

with torch.inference_mode():
    output = pipeline(**inputs)
    output.images[0].save("edited.png")

Multi-Image Composition

import torch
from PIL import Image
from diffusers import ZenArtistEditPlusPipeline

pipeline = ZenArtistEditPlusPipeline.from_pretrained(
    "zenlm/zen-artist",
    torch_dtype=torch.bfloat16
)
pipeline.to("cuda")

image1 = Image.open("person.jpg").convert("RGB")
image2 = Image.open("background.jpg").convert("RGB")

prompt = "Place the person naturally in the scenic background"

inputs = {
    "image": [image1, image2],
    "prompt": prompt,
    "generator": torch.manual_seed(42),
    "true_cfg_scale": 4.0,
    "negative_prompt": " ",
    "num_inference_steps": 40,
    "guidance_scale": 1.0,
    "num_images_per_prompt": 1,
}

with torch.inference_mode():
    output = pipeline(**inputs)
    output.images[0].save("composed.png")

Multi-GPU Deployment

export NUM_GPUS_TO_USE=4
export TASK_QUEUE_SIZE=100
export TASK_TIMEOUT=300

Hardware Requirements

Setup	VRAM	Notes
Minimum (FP8)	8GB	Reduced quality
Recommended (BF16)	24GB	Full quality
Optimal (Multi-GPU)	4x 24GB	High throughput

Community Support

Zen Artist is compatible with the HuggingFace diffusers ecosystem, including LoRA workflows, FP8 quantization, and layer-by-layer offloading for low-VRAM deployments.

License

Apache 2.0

Citation

@misc{zenlm2025zen-artist,
    title={Zen Artist: Diffusion-based Image Generation and Editing},
    author={Hanzo AI and Zoo Labs Foundation},
    year={2025},
    publisher={HuggingFace},
    howpublished={\url{https://huggingface.co/zenlm/zen-artist}}
}

---

Zen LM by Hanzo AI - Clarity Through Intelligence
zenlm.org  |  HuggingFace  |  GitHub