Ideogram4Scheduler - ComfyUI Built-in Node Documentation - ComfyUI

The Ideogram 4 Scheduler node generates a sequence of sigma values (noise levels) for the diffusion sampling process, based on the Ideogram 4 reference schedule. It creates a custom noise schedule that adapts to the image dimensions and allows fine-tuning through statistical parameters.

Inputs

Parameter	Description	Data Type	Required	Range
`steps`	The number of sampling steps to generate the schedule for (default: 20). The output contains `steps + 1` sigma values.	INT	Yes	1 to 200
`width`	The width of the image in pixels (default: 1024). The resolution relative to a 512×512 reference shifts the noise schedule.	INT	Yes	256 to 8192 (step: 16)
`height`	The height of the image in pixels (default: 1024). The resolution relative to a 512×512 reference shifts the noise schedule.	INT	Yes	256 to 8192 (step: 16)
`mu`	The mean parameter for the logit-normal distribution, controlling the central noise level. Combined with the resolution term to form the logSNR shift (default: 0.0).	FLOAT	Yes	-10.0 to 10.0 (step: 0.05)
`std`	The standard deviation parameter for the logit-normal distribution, controlling the spread of noise levels (default: 1.75).	FLOAT	Yes	0.1 to 5.0 (step: 0.05)

Note: The schedule is derived from a logit-normal distribution over reference time. A resolution term equal to 0.5 * log((width × height) / (512 × 512)) is added to mu, so larger or smaller images shift the schedule relative to a 512×512 reference at the same mu value.

Outputs

Output Name	Description	Data Type
`SIGMAS`	A tensor of sigma values representing the noise schedule, with length equal to `steps + 1`. The values descend from high noise to low noise, with the final value set to 0.0 for complete denoising.	SIGMAS

This documentation was AI-generated. If you find any errors or have suggestions for improvement, please feel free to contribute! Edit on GitHub

Source fingerprint (SHA-256): af0749713ce223d2246fc24b5100f18aa68d56746480990282899c223578b8f4