Dynamic Forms
When you select a Pipe in Prompt Studio, the form that appears isn't a hardcoded set of fields — it's generated dynamically from the underlying ComfyUI workflow. Sweet Tea reads the workflow's node graph and runtime metadata, figures out what parameters exist, and builds a usable form out of them. This is how arbitrary workflows become point-and-click tools.
How Forms Are Built
Every Pipe has a schema derived from its workflow graph and ComfyUI's object_info. The schema defines what fields to show, what types they are, and how they're organized. When you select a Pipe, Studio reads this schema and renders the form.
This means two things: first, every ComfyUI workflow can have a form (not just pre-built templates). Second, the form's quality depends on how well the Pipe's schema has been configured. A freshly imported Pipe might show every raw parameter name; a well-curated Pipe shows clean labels and sensible groupings.
Tip: If a form feels cluttered or confusing, the fix is in the Pipe's schema, not the workflow itself. See Advanced Pipe Management for how to rename fields, hide low-value controls, and reorder the layout.
Core vs. Advanced Controls
Forms split their fields into two tiers to keep your day-to-day workflow clean:
Core controls
These are the fields you'll touch on almost every run — the high-signal parameters that directly shape your output:
- Prompt and negative prompt — What you want (and don't want) in the image
- Resolution / dimensions — Output width and height
- Steps — How many denoising steps the sampler runs (more steps = more detail, up to a point)
- CFG scale — How strongly the image follows your prompt vs. creative freedom
- Seed — Controls randomness. Same seed + same settings = same image
Advanced controls
These are the deeper knobs you'll reach for when fine-tuning or diagnosing specific behavior:
- Node-specific modifiers and thresholds
- Sampler and scheduler selection
- Bypass toggles for optional pipeline stages
- Model-specific parameters
You can expand the Advanced section anytime. It's collapsed by default to keep the form focused.
Field Types
The form renders different control widgets depending on the parameter type:
| Field type | What it looks like | Typical examples |
|---|---|---|
| Text | A text input or textarea | Prompt, negative prompt, caption fields |
| Dropdown | A select menu with fixed options | Model selector, sampler name, scheduler type |
| Number / Slider | A numeric input with optional slider | Steps (10-50), CFG scale (1.0-20.0), denoise strength (0.0-1.0) |
| Toggle | An on/off switch | Enable high-res fix, bypass upscaling stage |
| Media input | A drop zone for images | Source image for img2img, mask for inpainting, control image |
Each field may show validation constraints — minimum/maximum values, required status, or type expectations. If a field turns red or shows an error, it means the current value won't work for generation.
Here are some typical value ranges you'll encounter:
| Parameter | Common range | What it controls |
|---|---|---|
| Steps | 10-50 | Denoising iterations. More steps = more detail, diminishing returns past ~30 |
| CFG Scale | 1.0-20.0 | Prompt adherence. Higher = stricter following, lower = more creative freedom |
| Denoise Strength | 0.0-1.0 | How much to change from the input image (img2img). 1.0 = fully regenerate |
| Width / Height | 512-2048 | Output resolution. Higher = more VRAM usage |
Note: Dropdown options (like available models or samplers) come from ComfyUI's runtime. If a model you installed doesn't show up in a dropdown, the engine may need a restart to pick up new files.
Node Bypass
Some Pipes expose bypass toggles that let you skip entire stages of the workflow. For example, a Pipe might include an upscaling step that you can toggle off when you're doing quick iterations and don't need high-res output yet.
Bypass is powerful for experimentation. Say you're working with a Pipe that has a generation stage, a face-fix stage, and an upscale stage. During early prompt exploration, you can bypass the upscale stage to get results faster, then enable it once you've found a prompt you like.
Warning: Bypassing a node can affect downstream values. If you bypass a stage and the output looks unexpected, re-enable it and compare before assuming something else is broken.
Form Persistence
Your form state is automatically saved as you work. If you navigate away from Prompt Studio and come back, your settings should be right where you left them — the same prompt, the same slider values, everything.
This persistence works across route changes within the app and even survives restarts in most cases. It's designed so you don't lose an in-progress prompt configuration because you needed to check something in the Gallery.
When you want to intentionally start fresh, reset the form to defaults. And when you've found a configuration that works well, save it to the Prompt Library so you can recall it anytime.
Tip: Save a "baseline" preset for each Pipe you use regularly. That way you always have a known-good starting point to return to after experiments.