Illustrious Prompt Result Viewer
A pass-through viewer for the DeepSeek prompt generator
- positive_prompt
- positive_prompt_chinese
- negative_prompt
- raw_response
What it is
The DeepSeek Illustrious prompt generator is a black box until you look inside: it returns four strings and you can't see any of them on the canvas. Illustrious Prompt Result Viewer is the pack's answer - an output node that shows all four on the node itself in one combined panel: the English positive prompt, the Chinese translation, the negative prompt, and the raw API response. Wire the generator's outputs into it and you finally get to read what the model actually wrote before you commit it to CLIP.
Why you'd reach for it
Two real reasons. First, trust but verify: the model runs a "thinking" pass, and sometimes the JSON it returns has a tag you didn't ask for, a weird weight, or content that got truncated. The raw_response pane - which includes the reasoning text when thinking mode is on - is where you spot that. Second, the workflow keeps the text: because the viewer stores what it displayed, re-opening a saved workflow shows the prompts that produced a given image. If you're iterating on prompts, that's genuinely handy. The README even marks it optional in the recommended chain, so think of it as an inspection window rather than a required step.
How it works
It's a pass-through with a display. The node takes four inputs (positive_prompt, positive_prompt_chinese, negative_prompt, raw_response), concatenates them into one labeled preview ([Positive Prompt] … [Raw Response]), and returns all four strings unchanged. Because it's declared an output node and forces itself to re-run every time, the preview is always fresh after an execution - but that cuts both ways, see below.
The one trap
It's an output node, and ComfyUI only executes nodes that lead to an output node. That's usually harmless - your workflow already ends in a Save Image, and the viewer just adds a second terminal that shows the generator's text. But if you wire the viewer in and the sampler branch stops feeding any output node, ComfyUI will happily prune it and you'll get your prompt panel and no image. The safe habit: keep the normal chain (generator → Dual Prompt CLIP Encode → KSampler → VAE Decode → Save Image) intact, and hang the viewer off the generator as a side terminal, or skip it once you trust the generator. Also, don't expect live updates while the API is still streaming - the panel refreshes after a run completes, not during it. Same install as the rest of the pack: no extra dependencies, nothing to download.
Inputs (4)
| Name | Type | Default | Description |
|---|---|---|---|
| positive_prompt | STRING | — | |
| positive_prompt_chinese | STRING | — | |
| negative_prompt | STRING | — | |
| raw_response | STRING | — |
Outputs (4)
| Name | Type | Description |
|---|---|---|
| positive_prompt | STRING | — |
| positive_prompt_chinese | STRING | — |
| negative_prompt | STRING | — |
| raw_response | STRING | — |