DP Sampler With Info

This is the stripped-down sibling of DP Advanced Sampler: a normal KSampler that, alongside the latent, spits out a plain-text summary of the settings it just ran. No built-in upscaling, no tiling - just sample, plus a receipt. If you've ever generated fifty images, found the good one, and then had no idea what steps and CFG produced it, this is the node that fixes that.

It's one of 105+ nodes in DesertPixelAi's Desert Pixel pack. Nothing exotic here - it's a quality-of-life wrapper around sampling - but it's a genuinely useful habit to build into a graph.

How it works

The sampling half is stock ComfyUI. Feed it a model, the latent_image, your positive and negative conditioning, and set seed, steps, cfg, sampler_name, scheduler, and denoise. It denoises and returns a LATENT exactly like a KSampler would - you still wire that into a VAE Decode to get an image.

The difference is the second output, sampler_info: a string listing the values it used. The point is to carry your parameters out of the node as data you can do something with, rather than leaving them buried in the node's widgets where a batch run overwrites them.

The inputs and outputs that matter

The knobs are the same four everyone touches: steps, cfg, the sampler_name/scheduler pair, and denoise (drop it below 1.0 for img2img or a hires second pass).

Two outputs. LATENT goes to your VAE Decode. sampler_info (STRING) is the interesting one - route it into a text-preview node to read it, or into a save node's caption/filename so the settings get stamped onto the output. That's the whole reason to reach for this over a plain KSampler.

Note what's not here: there's no vae input and no image output. Unlike DP Advanced Sampler, this node stops at the latent. If you want the built-in hires-fix and tiling, that's the Advanced one; this is the lean version.

How to install it

Via ComfyUI Manager: search Desert Pixel, install, restart. Or manually:

cd ComfyUI/custom_nodes
git clone https://github.com/DesertPixelAi/ComfyUI-Desert-Pixel-Nodes

Restart ComfyUI, pip install -r requirements.txt if the repo ships one. No models to download; it runs on your existing checkpoint. Find it under the DP category.

Common issues & troubleshooting

Same sampler rules as any KSampler. The sampler_info output doesn't change what a good setting is. DPM++ 2M Karras remains the default for SD 1.5 and SDXL; flow-matching models (Flux, Z-Image) want Euler-family on a conservative scheduler like beta or simple, and will fall apart on Karras. The info string reports what you chose - it can't tell you it was a bad choice.

"Where's my image?" There isn't one - the node outputs a latent, not a picture. Add a VAE Decode after it. If you want decode built in, use DP Advanced Sampler instead.

The info text isn't showing up. It's an output, not a display. Wire sampler_info into a text preview or a save-image caption field; on its own it does nothing visible.