| Property | Value |
|---|---|
| License | Non-commercial |
| Language | English |
| Pipeline | Text-to-Image |
| Format | Safetensors |
What is FLUX.1-schnell-dev-merged-fp8?
FLUX.1-schnell-dev-merged-fp8 is an optimized text-to-image model that combines the capabilities of FLUX.1-dev and FLUX.1-schnell architectures in a merged FP8 format. This model is specifically designed for efficient inference with minimal steps while maintaining quality output.
Implementation Details
The model utilizes a specialized architecture optimized for fast inference, supporting various schedulers including Normal, sgm_uniform, simple, and beta. It's implemented as a .safetensors model that should be placed in the "models\unet" directory for ComfyUI compatibility.
- Recommended inference steps: 6-8
- Optimal guidance scale: 3.5
- Primary sampler: Euler
- Supports dual-pass generation for refined results
Core Capabilities
- Fast image generation with 6-8 inference steps
- Two-pass refinement system
- Multiple scheduler compatibility
- ComfyUI workflow integration
- FP8 precision for optimal storage efficiency
Frequently Asked Questions
Q: What makes this model unique?
This model stands out for its efficient dual-pass generation system and FP8 precision, allowing for high-quality image generation in just 6-8 steps. The merged architecture combines the strengths of both FLUX.1-dev and FLUX.1-schnell models.
Q: What are the recommended use cases?
The model is ideal for rapid prototyping and generation of images where speed is crucial. It's particularly effective when used with the recommended two-pass system: 6 steps first pass with 2-3 steps second pass, or 8 steps first pass with 2 steps second pass for refinement.