If you’ve been looking for a lightweight yet powerful anime AI model, Anima Preview 2 might be exactly what you need.
In a space dominated by heavy models requiring high-end GPUs, this one stands out by delivering high-quality anime-style images while running on low VRAM hardware—making it accessible to almost everyone.
🎬 Watch the Full Workflow in Action
🚀 What is Anima Preview 2?
Anima Preview 2 is a 2 billion parameter text-to-image model specifically designed for anime-style generation.
Instead of chasing realism, it focuses entirely on:
- 🎭 Expressive characters
- 🎨 Stylized visuals
- ✨ Clean anime aesthetics
This makes it ideal for:
- Character design
- Anime fan art
- Stylized illustrations
- YouTube thumbnails & content creation
💻 Runs on Just 6GB VRAM
One of the most impressive aspects of this model is its efficiency.
👉 At ~4.18 GB in size, it can run on:
- Older GPUs
- Mid-range systems
- Even entry-level setups
And despite that, it delivers:
- ⚡ Fast generation speed
- 🔥 High-quality outputs
- 🧠 Consistent character styling
You don’t need a high-end machine to start creating stunning anime visuals.
🎯 Control Style Like Never Before
This is where Anima Preview 2 really shines.
You can guide the output using:
- 🎨 Artist names
- 📺 Anime series
- 🕰️ Specific eras or styles
👉 Guide it using artists, anime, or eras.
This gives you fine control over the visual direction, allowing you to:
- Recreate classic anime vibes
- Generate modern cinematic anime scenes
- Blend styles creatively
🧰 ComfyUI Workflow Breakdown
In the video, I walk you through a complete ComfyUI workflow designed for this model.
🔹 Models Used:
- Anima Preview 2
- Qwen 3.0 6B Text Encoder
- Qwen Image VAE
🔹 Upscaling:
- 4x-AnimeSharp.pth
⚙️ Workflow Sections:
1. Models Section
Load your:
- Main model
- Text encoder
- VAE
2. Prompt Section
Define:
- Positive prompt (detailed description)
- Negative prompt (to avoid artifacts)
3. Sampling Section
This is where:
- The actual image generation happens
- You control steps, CFG, and sampler
4. Upscaling Section
Since this is a preview model:
- Max base resolution: ~1K
But with upscaling:
- ✅ 4K
- ✅ 8K
- ✅ Even 16K output
🧪 Sample Prompt Styles
Here are some styles tested in the video:
- Naoki Urasawa-inspired realism
- Akira-style cyberpunk scenes
- Gantz sci-fi aesthetic
- Classic anime styles (Naruto, Evangelion, etc.)
- Modern cinematic anime lighting
The model handles all of these surprisingly well, especially considering its size.
🔥 Why This Model Stands Out
✔️ Low VRAM requirement
✔️ Fast performance
✔️ Strong anime stylization
✔️ Flexible style control
✔️ Works great with ComfyUI
🙌 Final Thoughts
Anima Preview 2 is a great example of how AI models are becoming:
- More efficient
- More accessible
- More specialized
If you’re into anime generation and want something that:
- Doesn’t require expensive hardware
- Still delivers great results
👉 This model is absolutely worth trying.