Which settings appear, their allowed values, and their defaults vary by model. A node’s config panel is the source of truth for the model you’ve selected — not every model exposes every setting below, and some are only available on specific tiers.
Format and composition
Controls for the shape, size, and encoding of the output.Width-to-height ratio of the output, such as
1:1 or 16:9.Output resolution, often expressed as a preset like
720p or 1080p.For aspect-ratio changes, the ratio the input is reframed to.
For upscaling, the resolution to produce, such as
4k.File format of the output, such as
png or jpeg.Quality preset that trades render time and cost against fidelity.
Whether the output background is opaque or transparent.
How closely an edit preserves the original input image.
Prompt controls
Settings that change how the model interprets and expands your prompt.Describes what to avoid in the output.
Lets the model rewrite your prompt for better results before generating.
Expands a short prompt into a more detailed one automatically.
Alternate name some models use for automatic prompt expansion.
Optimizes the prompt for the model’s preferred phrasing.
Allows the model to search the web for current context while generating.
Generation strength (image and video)
Controls for how literally the model follows the prompt and how much it transforms the input.How strictly the model adheres to the prompt. Higher values follow it more closely.
Classifier-free guidance scale — another name for prompt adherence strength on some models.
Number of denoising steps. More steps can add detail at the cost of speed.
How much the output departs from the input image. Higher values change it more.
How strongly reference images influence the result.
How strongly a style reference is applied.
Neutralizes color casts carried over from reference inputs.
Speed preset that trades render time against quality.
Video
Settings specific to video generation and editing.Length of the generated clip, in seconds.
Generates a matching audio track alongside the video.
Adds generated background music to the clip.
Locks the camera in place so the shot doesn’t move.
How much motion the model introduces into the scene.
Retains the source clip’s audio when editing video.
Preserves the original sound rather than replacing it.
How a character is framed or facing in the output.
Keeps the existing audio track when upscaling video.
Speed-versus-quality preset for the render.
Quality preset for the produced video.
Upscale (image)
Image upscalers expose detail controls that shape how much the model invents versus preserves. The most common ones:Upscale factor, such as
2 or 4.How much new detail the model may add. Higher values invent more.
How closely the result stays true to the original image.
How much fine, repeating texture detail is introduced.
Strength of high-dynamic-range enhancement.
Amount of sharpening applied to the result.
Applies dedicated restoration to faces. Some models add
face_enhancement_strength and face_enhancement_creativity to tune it.Tunes the upscaler for a content type, such as photos or art.
Selects the underlying upscaling engine when a model offers more than one.
Upscaling mode the model runs in.
Upscalers vary widely. Beyond these, individual models expose their own detail knobs — the config panel lists every control available for the model you pick.
Speech (text-to-speech)
Settings that shape synthesized voice.Consistency of the voice. Higher values are steadier; lower values are more expressive.
How closely the output matches the source voice.
How much stylistic exaggeration is applied to the delivery.
Speaking rate of the voice.
Music
Settings for music generation.Length of the generated track, in seconds.
Generates an instrumental track with no vocals.
Text and LLMs
Settings for text-generation models.How much the model reasons before answering. Higher effort can improve quality at the cost of speed.
Allows the model to search the web for current information.
Constrains the output to a structured format using a predefined schema.