NewBie-image-Exp0.1 - NewBieAI-Lab open source experimental anime literate graphical models

What is NewBie-image-Exp0.1?

NewBie-image-Exp0.1 is the first experimental anime text-generation model open-sourced by the NewBieAI-Lab team, which adopts the Next-DiT architecture with 3.5B parameters, optimized for the secondary genre. The model improves prompt comprehension through the fusion of dual text encoders (GEMMA3-4B-IT and Jina CLIP v2) and innovatively introduces XML structured prompts to enhance the generation stability of multi-character scenes. Highlights include 16-channel VAE to optimize detail performance, support for LoRA fine-tuning, and the ability to run with only 8GB of video memory, making it suitable for scenarios such as anime characterization and homage creation. Model to non-commercial license open source , weights and derivatives need to apply for a license , the source code follows the Apache 2.0 protocol .

Features of NewBie-image-Exp0.1

• Multiple model support: Supports a wide range of images, video, audio, and 3D models, including but not limited to SD1.x, SD2.x, SDXL, Pixart, AuraFlow, HunyuanDiT, Flux, Lumina Image 2.0, HiDream, Qwen Image, and more.
• asynchronous queuing system: Generation tasks can be processed asynchronously.
• Intelligent Memory Management: Run large models even on GPUs with as little as 1GB of VRAM.
• Support for multiple file formats: Can load ckpt and safetensors files, supports embedded/text reversal, LoRas, Hypernetworks and more.
• Workflow management: You can save and load workflows as JSON files, load workflows from generated PNG, WebP and FLAC files.

Core Advantages of NewBie-image-Exp0.1

• High quality image generation: Generate high-quality animation images with rich details and remarkable visual effects to meet the needs of professional creation.
• Strong architectural foundationThe advanced Next-DiT architecture, combined with a 3.5B parameter count, provides powerful image generation capabilities to support complex scenes and diverse styles.
• Efficient pre-training data: Pre-training using the large-scale full dan + 1m e621 dataset ensures that the model learns a rich set of image features and styles to enhance the diversity and accuracy of the generated images.
• Optimized hardware training: The model is trained for a long time on high-performance hardware, accumulating about 23,000 h200 hours to ensure the stability and efficiency of the model.
• Flexible Text Encoder: Combining Google/Gemma3-4b-it and Jina Ai/Jina Clip v2 to enhance the understanding of text descriptions for more accurate image generation.
• Wide range of applicability: Suitable for animation creation, game design, digital art, advertising and marketing, education and other fields, with a wide range of industry applications.
• Innovative VAE Configurations: Adoption Flux 1 Dev-VAE performs image encoding and decoding to optimize the quality and detail of the resulting image and enhance the overall visual effect.

What is NewBie-image-Exp0.1's official website?

• Github repository:: https://github.com/NewBieAI-Lab/NewBie-image-Exp0.1
• Huggingface Model Library:: https://huggingface.co/NewBie-AI/NewBie-image-Exp0.1

Applicable people of NewBie-image-Exp0.1

• Anime and Manga Artists: Creators who need to quickly generate high-quality anime characters and scenes for inspiration or directly as creative material.
• game developer: Teams that need to efficiently generate character, scene, and concept art in game design, especially for anime-style game projects.
• Digital art creators: Artists seeking a unique anime-style digital artwork that can be used to create illustrations, concept art, etc.
• Cartoonists and illustrators: Creators who need to generate sketches and illustrations quickly for projects such as comic book creation, book illustration, and more.
• Advertising and marketing staff: Professionals who need appealing anime style images for advertising and social media promotion.
• Film & Animation Production Team: Accelerate the creative development process for teams that need to generate concept art and subplots in film, television and animation pre-production.