Skip to main content
Parameter Referenceguidance_scaleFLUX.1 Dev

FLUX Guidance Scale

Comprehensive Calibration Guide, Syntax & Recommended Values

Calibrates flow-matching transformer guidance strength in FLUX.1 models. Balances prompt compliance against natural contrast and skin tone fidelity.

Parameter Flag
guidance_scale
Default Setting
3.5
Allowed Range
1.0 to 10.0 (Optimal: 2.5 to 4.0)
Target Architecture
FLUX.1 Dev
Standard Command Syntax
guidance_scale: <1.0 - 10.0>

How FLUX Guidance Scale Operates Internally

Unlike traditional diffusion models that use Classifier-Free Guidance (CFG), FLUX.1 Dev uses flow-matching trajectory guidance. In FLUX, values between 2.5 and 3.5 produce the most natural photographic skin tones and organic depth. Exceeding 4.5 frequently introduces high contrast, harsh edge burn, and plastic skin smoothing artifacts.

Recommended Value Calibration Matrix

Setting / SyntaxTarget ScenarioVisual Result & Impact
guidance_scale: 2.5Ultra-Organic PhotorealismSoft natural contrast, gentle film grain, highly authentic skin tones.
guidance_scale: 3.5Standard FLUX Sweet SpotOptimal prompt adherence, sharp typographical rendering, balanced dynamic range.
guidance_scale: 5.0+Complex Typography & SignageEnforces literal text compliance on signs and logos at the expense of slight saturation boost.

Common Mistakes & Syntax Errors

  • •Using SDXL-style guidance values (7.0 - 9.0) in FLUX, which severely burns colors and produces plastic artifacts.
  • •Trying to use negative prompts to fix guidance issues—FLUX ignores negative text natively.

Production Prompt Examples

RAW street photograph of an antique store with a sign reading "TIME TRAVEL VINTAGE", shot on 35mm film, natural light, guidance_scale: 3.0

Ready to experiment with guidance_scale?

Launch the Visual Prompt Compiler to test real-world camera optics, lighting, and parameters in real time.

Test in Compiler Studio

FLUX Guidance Scale Frequently Asked Questions

Why does FLUX guidance scale differ from Stable Diffusion CFG?

FLUX uses a rectified flow transformer rather than standard DDPM/U-Net latent diffusion. The mathematical velocity vectors require lower numerical scale coefficients (2.5–3.5 vs SDXL's 6.0–8.0).