Skip to content

Generation Parameters

Every model in OpenCauldron supports a different set of parameters. The interface adapts automatically — you’ll only see the parameters that apply to your selected model.

These parameters appear for most or all models.

The width-to-height ratio of the generated output. Available for all models.

Common ratios and their typical use cases:

RatioUse case
1:1Social media posts, profile images
16:9Landscape, presentations, desktop wallpapers
9:16Stories, Reels, TikTok, mobile wallpapers
4:3Traditional photo, print
3:4Portrait photography, posters

Some models support additional ratios like 2:1, 3:2, 21:9, and more. The available options update based on your selected model.

A number that controls the random generation process. Available on most image models and some video models.

  • Leave empty for a random result each time
  • Use the same seed with the same prompt and settings to reproduce an identical result
  • Useful for iterating on a composition — change the prompt slightly while keeping the seed fixed

Choose between PNG (lossless, larger files) and JPEG (compressed, smaller files). Available on Flux, Imagen, and Gemini models.

Use PNG when you need transparency or will edit the image further. Use JPEG for faster downloads and sharing.


Describe what you want to exclude from the generated image. The model will actively try to avoid these elements.

Available on: Imagen 4, Gemini Flash, Ideogram 3, Recraft, Kling, Veo

Examples:

  • blurry, low quality, watermark — general quality control
  • text, letters, words — avoid unwanted text
  • people, faces, hands — exclude human figures

Controls the output image dimensions. Higher resolution means more detail but slower generation.

ModelOptions
Imagen 41K, 2K
Gemini Flash / Lite512, 1K, 2K, 4K
Grok Imagine1K, 2K

Apply a predefined artistic style to the output. Available on Ideogram 3 and Recraft.

Ideogram styles include realistic, cinematic, anime, watercolor, sketch, 3D, and more. Recraft offers realistic image, digital illustration, vector illustration, and icon styles with additional sub-style options.

Trade off between speed and quality. Available on Ideogram 3.

OptionDescription
TurboFastest generation, slightly lower detail
DefaultBalanced speed and quality
QualitySlowest, highest detail and coherence

Controls how strictly the model follows your prompt. Available on Flux Dev.

  • Low values (1.5–2.5) — more creative, looser interpretation
  • High values (3.5–5.0) — more literal, closely follows the prompt
  • Default: 3.0

Number of diffusion steps during generation. Available on Flux Dev.

  • Fewer steps (1–15) — faster, rougher output
  • More steps (25–50) — slower, finer detail
  • Default: 28

Classifier-free guidance scale — similar to Guidance but used by different models. Available on Recraft (as artistic level).

Controls the balance between prompt adherence and creative freedom. Range: 0.3–0.7 (default: 0.5).

Controls whether the model can generate images of people. Available on Imagen 4 and Gemini models.

OptionDescription
Allow AllGenerate people of any age
Allow AdultsOnly generate adults (default)
Don’t AllowNo people in the output

This is a Google safety requirement. If your prompt mentions people and this is set to “Don’t Allow,” the generation may fail.


Length of the generated video clip in seconds. Available options depend on the model:

ModelDurations
Veo 35s, 8s
Veo 3.14s, 6s, 8s
Runway Gen-45s, 10s
Kling 2.15s, 10s
Hailuo 2.36s, 10s
Ray 25s, 10s
Ray Flash 25s, 9s

When enabled, the model produces a synchronized audio track with the video. Available on Veo 3/3.1 and Hailuo 2.3.

Audio is generated natively — the model creates sounds that match the visual content (footsteps, speech, ambient noise, etc.).

Apply a predefined camera movement to the video. Available on Ray 2 and Ray Flash 2.

OptionDescription
None (auto)Model decides camera behavior
Pan Left / RightHorizontal camera slide
Zoom In / OutMove toward or away from subject
Orbit Left / RightCircle around the subject

These are on/off switches that appear at the bottom of the parameters panel.

Adds a provider watermark to the output. On by default. Available on Imagen 4 and Gemini models.

Disabling the watermark may affect your usage terms with the provider — check your API agreement.

Lets the AI provider automatically rewrite your prompt before generation to improve results. Available on Imagen 4 and Gemini models.

When enabled, the provider may add detail, improve grammar, or restructure your prompt. The original prompt is preserved — enhancement happens behind the scenes.

Similar to Provider Prompt Enhance, but used by Hailuo/MiniMax. Optimizes your prompt on the provider side for better video generation results.

Makes the generated video seamlessly loop back to its first frame. Available on Ray 2 and Ray Flash 2.

Useful for creating background animations, loading screens, or social media content that plays on repeat.


Quick reference for which image models support which parameters:

ParameterImagen 4Gemini FlashFlux ProFlux DevIdeogram 3RecraftGrok
Negative PromptYesYesYesYes
ResolutionYesYesYes
SeedYesYesYesYesYes
Output FormatYesYesYesYes
StyleYesYes
Rendering SpeedYes
GuidanceYes
StepsYes
CFG ScaleYes
Person GenerationYesYes
WatermarkYesYes
Prompt EnhanceYesYes