MistoLine_Flux.dev
| Property | Value |
|---|---|
| Base Model | FLUX.1-dev |
| License | flux-1-dev-non-commercial-license |
| Parameters | 1.4B |
| Type | ControlNet for Text-to-Image |
What is MistoLine_Flux.dev?
MistoLine_Flux.dev is an advanced ControlNet model specifically designed for processing lineart and outline sketches. Built on the Flux1.dev architecture, it employs a sophisticated dual-stream Transformer structure that enhances alignment between image conditions and text prompts. The model demonstrates exceptional versatility across various applications, from industrial design to animation.
Implementation Details
The model architecture features a scalable Transformer module with dual-stream processing, optimized for both T5 and clip-l TextEncoders. It supports multiple precision formats including fp16 and fp8, with varying trade-offs between quality and speed. Training was conducted using A100-80GB GPUs with bf16 mixed precision.
- Supports high-resolution processing (720px+ on short edge)
- Controllable strength parameters (0.6-0.85 recommended)
- Guidance scale range: 3.0-5.0
- Minimum 30 steps recommended for optimal results
Core Capabilities
- Advanced lineart and outline processing
- Multi-domain compatibility (industrial, architectural, artistic)
- Balanced performance between image conditions and text prompts
- Flexible precision options for different performance needs
Frequently Asked Questions
Q: What makes this model unique?
The model's dual-stream Transformer architecture and extensive training on diverse scenarios make it particularly effective for lineart processing while maintaining balanced performance between image and text inputs.
Q: What are the recommended use cases?
The model excels in industrial design, architecture, interior design, animation, games, and photography applications, particularly when precise control over lineart and outlines is required.
