Z-Image-Turbo Anime
่ฉณ็ดฐ
ใใกใคใซใใใฆใณใญใผใ (1)
ใใฎใใผใธใงใณใซใคใใฆ
ใขใใซ่ชฌๆ
โ# ๐จ Z-Image-Turbo-Unleashed | Enhanced Anime & 3D Generation
Custom LoRA Merge โข Enhanced Anime Style โข 3D Style Support โข FP8
---
## โจ What is Z-Image-Turbo-Unleashed?
Z-Image-Turbo-Unleashed is a custom anime checkpoint built by merging carefully selected and purpose-trained LoRAs into the Z-Image-Turbo base model.
This merge significantly enhances anime aesthetics and can produce 3D-style images, while maintaining excellent generation quality across all use cases.
All integrated LoRAs were specifically trained and fine-tuned using a curated and enhanced mixed dataset, resulting in a model that handles a wide range of anime styles and creative concepts with precision โ no extra LoRAs or workarounds needed.
---
### ๐ก Z-Image-Turbo-Unleashed โ FP8
Lighter weight, fast, enhanced!
* โ Smaller file size โ easier to download and store
* โ Enhanced and refined dataset
* โ Can produce 3D-style images
โ Works great at just *8 steps** with Euler Ancestral
---
### โจ Key Features
โก *8-step turbo mode** โ Lightning fast with Euler Ancestral
๐จ *Enhanced anime style** โ Custom multi-LoRA merge for superior aesthetics
๐ง *3D style support** โ Can generate images with a 3D render look
๐ *Enhanced dataset** โ Refined and enriched for better overall quality
โญ *Euler Ancestral + Simple** โ Optimized sampler combo
---
## ๐ฏ Quick Start
### Installation
1. Download Z-Image-Turbo-Unleashed-FP8
2. Place the file in:
ComfyUI/models/checkpoints/
3. Load it using the Load Checkpoint node
4. Generate ๐
### โก Recommended Settings
```
Steps: 8
CFG: 1.0
Sampler: euler_ancestral
Scheduler: simple
Resolution: 1024ร1024
```
---
## ๐ Test Results
---
### ๐ฌ Test 1: Anime Portrait
Prompt:
```
A mysterious anime girl sitting by a rain-streaked window at night,
holding a warm cup of tea, wearing an oversized knit sweater. Soft
blue moonlight illuminating her face, melancholic yet peaceful expression.
Detailed interior with bookshelves and warm lamp light in the background.
```
Use Case: Atmospheric portraits, mood illustrations, visual novel art
---
### ๐ฌ Test 2: Fantasy Landscape
Prompt:
```
A hidden valley where enormous ancient trees with twisting roots
grow out of floating islands suspended above a crystal clear lake.
Waterfalls pour from the edges of each island, dissolving into fine
mist before reaching the water below. Bioluminescent flowers dot
the mossy ground, casting soft blue and violet light across the scene.
A stone bridge covered in ivy connects two islands, half crumbled by
time. The sky above is a deep twilight gradient of orange and deep
purple, with two moons visible through the canopy. Breathtaking
fantasy landscape, painterly detail, Studio Ghibli inspired atmosphere.
```
Use Case: Fantasy art, wallpapers, concept illustrations
---
### ๐ฌ Test 3: 3D Style Render (NEW)
Prompt:
```
A mysterious woman with long white hair and piercing violet eyes
standing inside an abandoned greenhouse overtaken by glowing plants,
rendered in detailed 3D anime style. She wears a fitted dark purple
coat with silver clasps over a high-collared black dress. One hand
gently touches a luminous flower that reacts to her presence, its
petals slowly opening and releasing tiny particles of golden light.
Vines with bioluminescent leaves crawl up rusted iron frames and
shattered glass panels around her. Moonlight pours in through the
broken ceiling. 3D anime CGI aesthetic, masterpiece quality.
```
Use Case: 3D anime art, CGI renders, key visuals
---
### ๐ฌ Test 4: Cyberpunk Cityscape
Prompt:
```
A vast cyberpunk megacity seen from a rooftop at golden hour,
towering skyscrapers covered in holographic advertisements stretching
endlessly into a hazy horizon. Elevated railways weave between
buildings as flying vehicles leave faint light trails across the
amber sky. Below, narrow streets glow with neon reflections on wet
pavement. Massive clouds roll in from the distance, lit from beneath
by the city's electric glow. Cinematic wide angle composition,
intricate architectural detail, atmospheric depth, masterpiece
quality anime illustration.
```
Use Case: Wallpapers, background art, worldbuilding
---
## ๐ก Prompting Guide
### Natural Language Works Best!
Good Example:
```
โ A mysterious anime girl sitting by a rain-streaked window at night,
holding a warm cup of tea, wearing an oversized knit sweater. Soft
blue moonlight illuminating her face, melancholic yet peaceful expression.
Detailed interior with bookshelves and warm lamp light in the background.
```
Bad Example:
```
โ anime girl, purple hair, window, rain, tea, sweater, sad
```
### ๐ง 3D Style Tips
To get a 3D render look, add keywords like:
```
rendered in detailed 3D anime style, 3D anime CGI aesthetic,
high polygon detail, cinematic depth of field
```
---
### General Tips
Do:
* โ Use natural language descriptions
* โ Be detailed (100โ300 words optimal)
* โ Include lighting and mood
* โ Describe pose and expression
* โ Add atmosphere details
* โ Specify art style references
* โ Include quality tags (masterpiece, detailed)
Don't:
* โ Use only tag-style prompts
* โ Write very short prompts
* โ Include conflicting styles
* โ Overload with negative prompts (usually not needed)
---
## ๐จ What Makes This Different?
### vs Original Z-Image-Turbo
Original:
- ๐จ Style: Photorealistic
- โ๏ธ Best Sampler: res_multistep
- ๐ Best Scheduler: simple
- ๐ข CFG: 1.0
- ๐ข Steps: 4โ8
- ๐ NSFW: Limited
- ๐ง 3D Style: No
- ๐ฏ Use Case: Photos
Unleashed FP8:
- ๐จ Style: Anime / Illustration / 3D style
- โ๏ธ Best Sampler: euler_ancestral โญ
- ๐ Best Scheduler: simple โญ
- ๐ข CFG: 1.0
- ๐ข Steps: 8
- ๐ NSFW: Partial (mixed dataset)
- ๐ง 3D Style: Yes
- ๐ฏ Use Case: Anime art, 3D style renders (all types)
### Custom LoRA Merge Benefits
* ๐จ Enhanced anime face rendering
* โจ Better hair and eye details
* ๐ธ Improved anime aesthetic overall
* ๐ง Can produce 3D-style images
* ๐ Enhanced and refined dataset
* ๐ฆ Single checkpoint file
---
## ๐ง Installation Guide
### Step 1: Download
Z-Image-Turbo-Unleashed-FP8.safetensors
### Step 2: Place File
```
ComfyUI/models/checkpoints/
โโโ Z-Image-Turbo-Unleashed-FP8.safetensors
```
### Step 3: Load & Generate
1. Open ComfyUI
2. Use the Load Checkpoint node
3. Select Z-Image-Turbo-Unleashed-FP8
4. Set:
Steps: *8**
CFG: *1.0**
Sampler: *euler_ancestral**
Scheduler: *simple**
Resolution: *1024ร1024**
5. Write a detailed natural-language prompt
6. Generate ๐จ
---
## ๐ Credits & License
### Base Model
* Developer: Tongyi Lab (Alibaba Group)
* Architecture: Single-Stream Diffusion Transformer (6B parameters)
* Algorithm: Decoupled-DMD + DMDR
* License: Apache 2.0
---
### Unleashed Conversion
* Custom LoRAs: Created and trained specifically for this merge
* Dataset: Enhanced mixed dataset
* Focus: Anime aesthetics + 3D style + creative flexibility
* Format: FP8
---
### Resources
* Original Model:
<https://huggingface.co/Tongyi-MAI/Z-Image-Turbo>
* GitHub:
<https://github.com/Tongyi-MAI/Z-Image>
* ComfyUI Integration:
<https://huggingface.co/Comfy-Org/z_image_turbo>
---
## ๐ Version History
### v2.0 โ FP8 Release & Major Update
๐ก *FP8 version released** โ Lighter weight, faster downloads
๐ *Enhanced dataset** โ Refined and enriched for better quality
๐ง *3D style support** โ Can now produce 3D-style images
โก *8-step generation** โ Euler Ancestral + Simple / CFG 1.0
---
Download, load with "Load Checkpoint", and unleash your anime creativity! ๐จ