README.md · chfm/Z-Image-Turbo-AIO at main

Z-Image-Turbo-AIO / README.md

chfm

Duplicate from SeeSee21/Z-Image-Turbo-AIO

f1dcefd 27 days ago

preview code

raw

history blame contribute delete

12.9 kB

	---
	license: apache-2.0
	language:
	- en
	- zh
	base_model:
	- Tongyi-MAI/Z-Image-Turbo
	pipeline_tag: text-to-image
	tags:
	- text-to-image
	- image-generation
	- diffusion
	- comfyui
	- photorealistic
	- bilingual
	- chinese
	- english
	- 8-step
	- fast-generation
	---

	# 🚀 Z-Image-Turbo Collection

	Ultra-Fast 8-Step Generation • Photorealistic & Anime • All-in-One

	[![License](https://img.shields.io/badge/License-Apache%202.0-blue.svg)](https://opensource.org/licenses/Apache-2.0)
	[![ComfyUI](https://img.shields.io/badge/ComfyUI-Compatible-green.svg)](https://github.com/comfyanonymous/ComfyUI)

	---

	## 📦 Models

	\| Model \| Style \| FP8 \| FP16 \| BF16 \| Best Sampler \|
	\|-------\|-------\|-----\|------\|------\|--------------\|
	\| 📸 Z-Image-Turbo-AIO \| Photorealistic \| ~10GB \| ~20GB \| ~20GB \| res_multistep + simple \|
	\| 🎨 Z-Image-Turbo-Anime \| Anime \| ~10GB \| ~20GB \| ~20GB \| euler_ancestral + beta \|

	All models: 8-9 steps • CFG 1.0 • 8GB VRAM compatible

	---

	## 🔧 Workflows

	\| Workflow \| File \| Best For \|
	\|----------\|------\|----------\|
	\| 🚀 [Standard v1](#-standard-v10--v20) \| `ZIT-AIO-v1.0.json` \| Simple text-to-image \|
	\| 🚀 [Standard v2](#-standard-v10--v20) \| `ZIT-AIO-v2.0.json` \| Improved upscaler \|
	\| 🎮 [ControlNet](#-controlnet) \| `ZIT-AIO-Control.json` \| Guided generation \|
	\| 🎲 [Seed Variance](#-seed-variance-enhancer) \| `ZIT-AIO-Variance.json` \| Output diversity \|
	\| 🎬 [SeedVR2](#-seedvr2) \| `ZIT-AIO-SeedVR2.json` \| Pro upscaling \|
	\| 🌊 [Depth V3](#-depth-anything-v3) \| `ZIT-AIO-DepthV3.json` \| Depth control + 3D \|
	\| 🎨 [Anime](#-anime-workflow) \| `ZIT-AIO-Anime.json` \| Anime generation \|

	---

	## 📸 Z-Image-Turbo-AIO (Photorealistic)

	Alibaba Tongyi Lab's 6B parameter photorealistic generator with integrated VAE + Text Encoder.

	Features: ⚡ 8-step • 📖 Bilingual EN/CN • 📸 Professional quality

	```yaml
	Steps: 9 \| CFG: 1.0 \| Sampler: res_multistep \| Scheduler: simple
	```

	🖼️ CivitAI: [BF16](https://civitai.com/models/2173571?modelVersionId=2447693) \| [FP16](https://civitai.com/models/2173571?modelVersionId=2550362) \| [FP8](https://civitai.com/models/2173571?modelVersionId=2448013)

	---

	## 🎨 Z-Image-Turbo-Anime (NEW!)

	Custom anime checkpoint with merged LoRAs for enhanced anime aesthetics.

	Features: ⚡ 8-step • 🎨 Enhanced anime style • 🔓 Partially NSFW capable

	```yaml
	Steps: 8-9 \| CFG: 1.0 \| Sampler: euler_ancestral \| Scheduler: beta
	```

	🖼️ CivitAI: [BF16](https://civitai.com/models/2259646?modelVersionId=2543657) \| [FP16](https://civitai.com/models/2259646?modelVersionId=2550879) \| [FP8](https://civitai.com/models/2259646?modelVersionId=2544019)

	---

	## 📋 Workflow Details

	<details>
	<summary><b>🚀 Standard v1.0 & v2.0</b> — Simple text-to-image</summary>

	### Standard Workflow

	![Standard Workflow Title](images/workflows/v2-title.png)

	Simple text-to-image with upscaler and dual sampler options.

	v2.0 Improvements:
	- Enhanced upscaler with denoise control (0.4-0.6)
	- Dual sampler: `res_multistep` or `euler_ancestral`
	- Dual scheduler: `simple` or `beta`
	- `scale_by` parameter for output size

	Required Nodes: rgthree-comfy, comfyui_image_metadata_extension

	![Standard Workflow Screenshot](images/workflows/v2-screenshot.png)

	📥 Download: [ZIT-AIO-v1.0.json](workflows/ZIT-AIO-v1.0.json) \| [ZIT-AIO-v2.0.json](workflows/ZIT-AIO-v2.0.json)

	</details>

	<details>
	<summary><b>🎮 ControlNet</b> — Guided generation with reference images</summary>

	### ControlNet Union Workflow

	![ControlNet Workflow Title](images/workflows/control-title.png)

	5 control types for precise composition control.

	Control Types:
	- Canny — Edge detection
	- HED — Soft edges
	- Depth — 3D depth
	- Pose — Body pose
	- MLSD — Line detection

	Settings:
	- ControlNet strength: 0.6-0.8
	- Megapixel scaling (auto aspect ratio)

	Required:
	- ControlNet Union file → `ComfyUI/models/model_patches/`
	- Download: [ControlNet Union](https://huggingface.co/alibaba-pai/Z-Image-Turbo-Fun-Controlnet-Union)

	Required Nodes: rgthree-comfy, comfyui_image_metadata_extension, comfyui_controlnet_aux

	![ControlNet Workflow Screenshot](images/workflows/control-screenshot.png)

	📥 Download: [ZIT-AIO-Control.json](workflows/ZIT-AIO-Control.json)

	</details>

	<details>
	<summary><b>🎲 Seed Variance Enhancer</b> — More diversity from same prompt</summary>

	### Seed Variance Enhancer Workflow

	![Seed Variance Workflow Title](images/workflows/variance-title.png)

	Adds controlled noise to text embeddings for more varied outputs.

	Settings:
	```yaml
	randomize_percent: 50
	strength: 20-30
	noise_insert: 'noise on beginning steps'
	steps_switchover_percent: 20
	```

	Trade-off: Higher values = more diversity but less prompt adherence

	Required Nodes: rgthree-comfy, comfyui_image_metadata_extension, [SeedVarianceEnhancer](https://github.com/ChangeTheConstants/SeedVarianceEnhancer)

	![Seed Variance Workflow Screenshot](images/workflows/variance-screenshot.png)

	📥 Download: [ZIT-AIO-Variance.json](workflows/ZIT-AIO-Variance.json)

	</details>

	<details>
	<summary><b>🎬 SeedVR2</b> — Professional diffusion upscaling</summary>

	### SeedVR2 Video Upscaler Workflow

	![SeedVR2 Workflow Title](images/workflows/seedvr2-title.png)

	DiT-based upscaling with superior quality and temporal consistency.

	Settings:
	```yaml
	resolution: 1536 (target short edge)
	batch_size: 4n+1 (1, 5, 9, 13, 17...)
	color_correction: 'lab'
	```

	Models: 3B (faster) or 7B (higher quality) • FP16/FP8/GGUF options

	Note: Models auto-download to `ComfyUI/models/SEEDVR2/`

	Required Nodes: rgthree-comfy, comfyui_image_metadata_extension, [SeedVR2](https://github.com/numz/ComfyUI-SeedVR2_VideoUpscaler)

	![SeedVR2 Workflow Screenshot](images/workflows/seedvr2-screenshot.png)

	📥 Download: [ZIT-AIO-SeedVR2.json](workflows/ZIT-AIO-SeedVR2.json)

	</details>

	<details>
	<summary><b>🌊 Depth Anything V3</b> — Depth control + 3D preview</summary>

	### Depth Anything V3 ControlNet Workflow

	![Depth V3 Workflow Title](images/workflows/depthv3-title.png)

	State-of-the-art depth-guided generation with dual modes.

	Two Modes:
	1. Generate — Depth-controlled image generation
	2. 3D Preview — View depth as 3D point cloud

	Models: da3_small / da3_base / da3_large / da3_giant
	- 8GB VRAM: Use up to `da3_large`

	Note: Models auto-download on first use

	Required Nodes: rgthree-comfy, comfyui_image_metadata_extension, [Depth Anything 3](https://github.com/ByteDance-Seed/Depth-Anything-3)

	![Depth V3 Workflow Screenshot](images/workflows/depthv3-screenshot.png)

	📥 Download: [ZIT-AIO-DepthV3.json](workflows/ZIT-AIO-DepthV3.json)

	</details>

	<details>
	<summary><b>🎨 Anime Workflow</b> — For Z-Image-Turbo-Anime</summary>

	### Z-Image-Turbo-Anime Workflow

	![Anime Workflow Title](images/workflows/anime-title.png)

	Optimized workflow for the anime checkpoint with integrated features.

	Includes:
	- Seed Variance Enhancer for diversity
	- SeedVR2 as alternative upscaler
	- Toggle switches for optional features
	- Image comparison nodes

	Best Settings:
	```yaml
	Sampler: euler_ancestral
	Scheduler: beta
	Resolution: 832×1216 (portrait) / 1216×832 (landscape)
	```

	Required Nodes: rgthree-comfy, comfyui_image_metadata_extension, SeedVarianceEnhancer, SeedVR2

	![Anime Workflow Screenshot](images/workflows/anime-screenshot.png)

	📥 Download: [ZIT-AIO-Anime.json](workflows/ZIT-AIO-Anime.json)

	</details>

	---

	## 🔄 FP8 vs FP16 vs BF16

	\| \| 🟡 FP8 (~10GB) \| 🔵 FP16 (~12GB) \| 🌟 BF16 (~20GB) \|
	\|---\|----------------\|-----------------\|-----------------\|
	\| Quality \| 95% \| 98% \| 100% \|
	\| Speed \| Fastest \| Fast \| Standard \|
	\| Download \| Fastest \| Medium \| Slowest \|
	\| Best For \| Daily use \| Balanced \| Final renders \|

	All versions work on 8GB VRAM!

	---

	## ⚙️ Common Settings

	```yaml
	# Essential (ALL workflows)
	Steps: 8-9
	CFG: 1.0 # Don't change!
	Negative Prompt: NONE (use ConditioningZeroOut)

	# Photorealistic
	Sampler: res_multistep
	Scheduler: simple

	# Anime
	Sampler: euler_ancestral
	Scheduler: beta
	```

	---

	## 💡 Prompting

	Natural language works best! (100-300 words)

	```
	✅ Professional food photography of artisan breakfast plate.
	Golden poached eggs on sourdough toast, crispy bacon, fresh
	avocado slices. Morning sunlight creating warm glow.

	❌ breakfast, eggs, toast, bacon, food photo, delicious
	```

	Bilingual Text Rendering:
	```
	English: Neon sign reading "COFFEE SHOP" in bright blue letters
	Chinese: Sign with "咖啡店" in elegant gold calligraphy
	```

	---

	## 📥 Downloads

	### Models — Z-Image-Turbo-AIO (Photorealistic)

	\| Precision \| CivitAI \| HuggingFace \|
	\|-----------\|---------\|-------------\|
	\| 🌟 BF16 \| [Download](https://civitai.com/models/2173571?modelVersionId=2447693) \| This page \|
	\| 🔵 FP16 \| [Download](https://civitai.com/models/2173571?modelVersionId=2550362) \| This page \|
	\| 🟡 FP8 \| [Download](https://civitai.com/models/2173571?modelVersionId=2448013) \| This page \|

	### Models — Z-Image-Turbo-Anime

	\| Precision \| CivitAI \| HuggingFace \|
	\|-----------\|---------\|-------------\|
	\| 🌟 BF16 \| [Download](https://civitai.com/models/2259646?modelVersionId=2543657) \| This page \|
	\| 🔵 FP16 \| [Download](https://civitai.com/models/2259646?modelVersionId=2550879) \| This page \|
	\| 🟡 FP8 \| [Download](https://civitai.com/models/2259646?modelVersionId=2544019) \| This page \|

	### Additional Downloads

	\| Source \| Link \|
	\|--------\|------\|
	\| 🎮 TensorArt \| [tensorart.me](https://tensorart.me/de-DE/models/939302608026244936/Z-Image-Turbo-AIO-FP8-FP8) \|

	### Additional Files

	\| File \| Link \|
	\|------\|------\|
	\| ControlNet Union \| [HuggingFace](https://huggingface.co/alibaba-pai/Z-Image-Turbo-Fun-Controlnet-Union) \|
	\| SeedVR2 Models \| Auto-downloads on first use \|
	\| Depth Anything V3 \| Auto-downloads on first use \|

	---

	## 📦 Custom Nodes

	<details>
	<summary><b>All Required Nodes</b></summary>

	Base (all workflows):
	- [rgthree-comfy](https://github.com/rgthree/rgthree-comfy)
	- [comfyui_image_metadata_extension](https://github.com/edelvarden/comfyui_image_metadata_extension)

	ControlNet Workflow:
	- [comfyui_controlnet_aux](https://github.com/Fannovel16/comfyui_controlnet_aux)

	Seed Variance / Anime Workflow:
	- [SeedVarianceEnhancer](https://github.com/ChangeTheConstants/SeedVarianceEnhancer)

	SeedVR2 / Anime Workflow:
	- [SeedVR2](https://github.com/numz/ComfyUI-SeedVR2_VideoUpscaler)

	Depth V3 Workflow:
	- [Depth Anything 3](https://github.com/ByteDance-Seed/Depth-Anything-3)

	</details>

	---

	## 📁 Files

	```
	├── z-image-turbo-fp8-aio.safetensors # 📸 Photorealistic FP8
	├── z-image-turbo-fp16-aio.safetensors # 📸 Photorealistic FP16
	├── z-image-turbo-bf16-aio.safetensors # 📸 Photorealistic BF16
	├── z-image-anime-aio-fp8.safetensors # 🎨 Anime FP8
	├── z-image-anime-aio-fp16.safetensors # 🎨 Anime FP16
	├── z-image-anime-aio-bf16.safetensors # 🎨 Anime BF16
	│
	├── workflows/
	│ ├── ZIT-AIO-v1.0.json # 🚀 Standard v1
	│ ├── ZIT-AIO-v2.0.json # 🚀 Standard v2
	│ ├── ZIT-AIO-Control.json # 🎮 ControlNet
	│ ├── ZIT-AIO-Variance.json # 🎲 Seed Variance
	│ ├── ZIT-AIO-SeedVR2.json # 🎬 SeedVR2
	│ ├── ZIT-AIO-DepthV3.json # 🌊 Depth V3
	│ └── ZIT-AIO-Anime.json # 🎨 Anime
	│
	├── images/
	│ ├── test1-coffee-shop.jpeg # Photo examples
	│ ├── test2-chinese-temple.jpeg
	│ ├── test3-sushi.jpeg
	│ └── workflows/ # Workflow images
	│ ├── v1-title.png & v1-screenshot.png
	│ ├── v2-title.png & v2-screenshot.png
	│ ├── control-title.png & control-screenshot.png
	│ ├── variance-title.png & variance-screenshot.png
	│ ├── seedvr2-title.png & seedvr2-screenshot.png
	│ ├── depthv3-title.png & depthv3-screenshot.png
	│ └── anime-title.png & anime-screenshot.png
	```

	---

	## 🙏 Credits

	Original Model: Tongyi Lab (Alibaba) • 6B S3-DiT • Apache 2.0

	AIO Conversion & Anime Merge: [SeeSee21](https://huggingface.co/SeeSee21)

	Resources:
	[Original HuggingFace](https://huggingface.co/Tongyi-MAI/Z-Image-Turbo) •
	[GitHub](https://github.com/Tongyi-MAI/Z-Image) •
	[ComfyUI Files](https://huggingface.co/Comfy-Org/z_image_turbo)

	---

	Download → Load Checkpoint → Generate! 🚀
	---

	<div align="center">

	Download, load with "Load Checkpoint", and generate professional photos in seconds! 🚀

	</div>