kkytai/scriptviz-flux-dev

FLUX.1-dev Handler (without PuLID)

High-quality image generation using FLUX.1-dev without PuLID character consistency.

Features

• FLUX.1-dev: Higher quality than schnell (28 steps vs 4)
• Lightweight: No PuLID dependencies, faster cold start
• Memory Efficient: Auto-detects GPU and applies appropriate offloading
• Batch Processing: Process multiple prompts in parallel

GPU Requirements

GPU	Mode	Batch Size	Notes
A100 80GB	Full GPU	4+	Best performance
L40S 48GB	CPU Offload	2-4	Good performance
A10G 24GB	Sequential Offload	1-2	Slower, but works

API Usage

Single Image Generation

{
  "inputs": "A professional portrait of a business person in a modern office",
  "parameters": {
    "num_inference_steps": 28,
    "guidance_scale": 3.5,
    "width": 1344,
    "height": 768
  }
}

Batch Generation

{
  "inputs": [
    "A sunset over mountains",
    "A city skyline at night",
    "A peaceful forest scene"
  ],
  "parameters": {
    "num_inference_steps": 28,
    "guidance_scale": 3.5
  }
}

Response Format

Single:

{
  "image": "data:image/png;base64,..."
}

Batch:

[
  {"image": "data:image/png;base64,..."},
  {"image": "data:image/png;base64,..."}
]

Parameters

Parameter	Default	Description
num_inference_steps	28	Denoising steps (20-50 recommended)
guidance_scale	3.5	Prompt adherence (3.0-7.0)
width	1344	Image width
height	768	Image height
seed	-1	Random seed (-1 for random)

Environment Variables

Variable	Default	Description
TORCH_COMPILE_MODE	reduce-overhead	Compile mode (reduce-overhead/default/false)
FIXED_BATCH_SIZE	4	Batch size for CUDA graphs
NUM_INFERENCE_STEPS	28	Default inference steps
ENABLE_OFFLOAD	auto	CPU offload (auto/true/false)

Deployment

1. Create a new model repository on Hugging Face Hub
2. Upload handler.py and requirements.txt
3. Create a Dedicated Endpoint (A100 80GB recommended)
4. Set Task type to "Custom"

Comparison with Other Handlers

Handler	Steps	Quality	PuLID	Cold Start	Use Case
flux-schnell	4	Good	No	Fast	Quick previews
flux-dev-no-pulid	28	Excellent	No	Medium	High quality generation
flux-dev-pulid	28	Excellent	Yes	Slow	Character consistency

When to Use This Handler

• You need high-quality image generation
• You don't need character consistency across images
• You want faster cold starts than the PuLID version
• You're debugging or testing without PuLID complexity

References

• FLUX.1-dev
• Diffusers FluxPipeline