ComfyUI image to prompt workflow from browser extraction to node-ready reverse prompt blocks.
Run one browser-first workflow to reverse-engineer a reference image into positive text, negative text, tags, and scene cues you can paste into ComfyUI and reuse on the second image.
ComfyUI Prompt Workbench
Upload one reference image and get positive prompt, negative prompt, and node-ready blocks for ComfyUI.
rabbit knight, rear view armor silhouette, consistent stance from Wonder3D multiview, copper rim light, hand-painted fantasy material detail, clean character sheet lighting
front-facing pose, mismatched armor, extra limbs, blurry back plate, washed out material, inconsistent character identity, low detail normal map artifacts
Positive CLIP Text Encode: subject + armor + lighting. Negative CLIP Text Encode: artifacts + identity drift. Control input: normal/depth/canny from the Wonder3D back view.
How to use in ComfyUI+
The shortest path is browser first, graph second. Extract the reverse prompt structure once, then decide inside ComfyUI which blocks deserve permanent node space and which should stay optional.
- Step 1: Upload a reference image in the browser. Start with a screenshot, render, inspiration board frame, or generated image. The workbench extracts subject, composition, lighting, and style before you touch the graph, so the prompt pack stays tied to the workflow you opened.
- Step 2: Review the positive, negative, and tag blocks. Keep the clean prompt as your main base, scan the negative block for cleanup terms, and mark the tags or scene cues you actually want to preserve.
- Step 3: Paste the sections into your ComfyUI text nodes. Move the clean prompt into your positive text node, send the negative block to your negative text node, and use the workflow handoff once the graph-ready split is worth saving.
- Step 4: Iterate inside the graph instead of re-describing the image. After the first pass, tune checkpoint, LoRA, sampler, CFG, or regional prompt choices in ComfyUI without reverse-engineering the whole image again.
What the output includes+
Searchers using the workflow modifier usually want something closer to a repeatable process than a one-off generator. This variant emphasizes graph-ready sections, faster reuse, easier handoff between browser extraction and node-based editing, and a cleaner second-image loop once the first pack works. It also covers the comfyui reverse prompt angle where the recovered prompt needs to survive actual node placement in positive and negative CLIP Text Encode blocks.
- Built for positive and negative separation before you move the prompt into a graph.
- Useful when the query is closer to comfyui reverse prompt than a generic image-to-prompt request.
- Useful for reference-image workflows, reverse-prompt cleanup, and fast first-pass extraction.
FAQ+
Is this a real ComfyUI workflow or only a prompt generator?
It is a browser-first workflow page. You upload a reference image, review structured reverse prompt sections, then move the positive, negative, and style blocks into your ComfyUI graph instead of starting from a loose paragraph.
Does this cover image to prompt comfyui workflow searches?
Yes. It is built for that query order as well as comfyui image to prompt workflow: extract the prompt pack in the browser, then map the positive, negative, and style sections into ComfyUI nodes.
What if I searched for comfyui reverse prompt?
That intent fits this page. The workflow is built for recovering a reusable prompt structure from a reference image, then splitting it into positive, negative, and optional style blocks you can keep tuning inside ComfyUI.
Which parts of the output go into positive and negative nodes?
Use the clean prompt or key tags in your positive text node, keep the negative block in the negative text node, and hold style or scene notes for later concat or manual tuning inside the graph.
Do I need custom nodes before I can use this workflow?
No. The first step happens in the browser. Once you get the prompt pack, you can paste the sections into the standard positive and negative text parts of your ComfyUI setup and then keep iterating with your usual nodes.
Related image-to-prompt workflows
Keep the ComfyUI route connected to nearby image to prompt, reverse prompt, and model-specific searches without blocking the workbench.