Two single-model workflows that do the same job with different prompt structures — and a side-by-side on why the prompt shape matters more than the sampler settings.
The idea
KJ-structured prompting treats the prompt as a filled-in template rather than free text: subject,
composition, light, material, style as separate fields. The model gets a consistent shape, which
reduces drift between generations.
Two routes
- FAST single model — optimized for turnaround, fewer steps, one model. Good for iteration.
- Instant single model — a second structural variant tuned for single-pass output.
They are not "fast vs quality" as such; they are two different prompt structures that ended up
sufficiently different to keep both.
What to watch for
Structured prompts help most when you are generating a series that has to stay consistent. For
one-off images the free-text prompt is often just as good and much faster to write.
Running Hub links
- https://www.runninghub.cn/post/2077705321124028417/?inviteCode=rh-v1111
- https://www.runninghub.cn/post/2078065107753590786/?inviteCode=rh-v1111
Files
Workflow JSONs and covers: https://huggingface.co/AI-KSK/ksk-video-278-comfyui-workflows
Video walkthrough (Chinese): https://youtu.be/g7HXab6m4Ys
Bilibili: https://www.bilibili.com/video/BV1BJNd6vE27/
Notes
- ComfyUI exports that reference models present in the cloud instance; not self-contained packs.
- Tutorial material, not a product.
- Independent AI-KSK release. Not affiliated with or endorsed by the model authors or RunningHub.
Top comments (0)