replicate/flux-3-klein-image-lora

Generate images from Flux 3 Klein Image using your own LoRA

Public
31 runs
License
GitHub
Weights
Playground API Examples README Versions

Examples

View more

Run time and cost

This model runs on Nvidia L40S GPU hardware. We don't yet have enough runs of this model to provide performance information.

Readme

FLUX 3 Image Klein

Generate an image from a text prompt. Choose the faster distilled mode or switch to the base model, and download the result as a PNG.

Getting Started

Enter a description of your image and leave the other settings at their defaults for a 1024x1024 image in distilled mode.

Example prompt:

Studio photograph of a cobalt-blue ceramic teapot on a brushed steel table, soft light from the left, a pale gray background, close-up framing with the entire teapot in view.

Describe the subject, setting, composition, lighting, and style you want. Put any text that should appear in the image in quotation marks. Prompt upsampling is on by default: it expands your description with scene, composition, and lighting details. Turn it off when you want your prompt used exactly as written.

Parameters

Input Default How It Works
prompt Required Describe the image you want to create. Include important details and where subjects should appear.
upsample_prompt true Expands your prompt into a more detailed description before generating. It adds processing time and can change the interpretation or composition. Turn it off to use your original wording. This changes the prompt, not the output resolution.
distilled true Uses the faster distilled mode. Turn it off to use the base model. The two modes can produce different results from the same prompt and seed.
width 1024 Output width in pixels.
height 1024 Output height in pixels.
num_inference_steps 0 Leave at 0 to use 16 steps in distilled mode or 50 in base mode. Set 1 to 100 to choose a step count yourself. Fewer steps are faster but can reduce image quality; more steps do not guarantee a better result.
guidance 3.5 Controls how strongly the base model follows your prompt, from 1 to 10. Higher values can emphasize the prompt but may make results look less natural. Ignored when distilled is on.
seed -1 Chooses a random seed by default. Use a nonnegative integer to repeat a result with the same prompt and settings on the same model version.

Image Size

Both dimensions must be multiples of 16, between 64 and 4096. The image must contain no more than 4,194,304 pixels in total.

Common choices include:

  • Square: 1024 x 1024.
  • Landscape: 1344 x 768.
  • Portrait: 768 x 1344.

Larger images generally take longer to generate. Changing the dimensions can also change the composition, even with the same prompt and seed.

Runtime and Pricing

This model runs on an NVIDIA L40S. Generation time depends on image size, mode, and step count. The first request after startup may take longer.

See Replicate’s pricing page for current pricing.

Model created
Model updated