vv

### Custom Node Testing

- [x] I have tried disabling custom nodes and the issue persists (see [how to disable custom nodes](https://docs.comfy.org/troubleshooting/custom-node-issues#step-1%3A-test-with-all-custom-nodes-disabled) if you need help)

### Expected Behavior

/

### Actual Behavior

@tobbez

### Steps to Reproduce

[``](url)

### Debug Logs

```powershell
My apologies for the echo in the system! I am fully locked in and ready to map this exact multi-stage showcase into a node structure.

To convert your fast-paced fashion timeline into a real generation, here is the explicit blueprint for wiring your nodes:

---

### Step 1: The Timeline Scheduler (The Brain)

Because your prompt switches clothes and scenes at exact timestamps (`[0:00-0:02]`, etc.), standard text boxes won't work. You need a node that understands time.

1. Add a **`BatchPromptSchedule`** node (found in custom node suites like *FizzNodes*).
2. Paste your entire timeline text block directly into it exactly as you wrote it.
3. Connect your **DualCLIPLoader** (the text encoders) directly into the `clip` input of this node.

---

### Step 2: Set the Video Boundaries

1. Add an **`EmptyFluxLatentVideo`** node (or a standard empty latent node if using a different base architecture).
2. Set your resolution to a vertical mobile format: **$720 \times 1280$**.
3. **The Math for Frame Count:** For a 7-second video targeted at 24 frames per second (fps), change your `frame_count` or batch size to exactly **168**.

---

### Step 3: Enforcing Clothes Consistency (The Secret Sauce)

Your prompt demands that outfits remain fully visible with *no missing items or replacement clothing*. To rigidly lock the style:

1. Add three individual **`Load Image`** nodes to your canvas.
2. Upload a standalone reference image of Outfit 1, Outfit 2, and Outfit 3 into each node respectively.
3. If you are using a state-of-the-art video model like [Seedance 2.0](https://comfy.org/workflows/064da31db8f3-064da31db8f3/), pass these image outputs into the model's specialized **Reference-to-Video** / IP-Adapter slots. This forces the engine to pull details directly from your images rather than guessing what a "cream fuzzy knit sweater" looks like frame-by-frame.

---

### Step 4: The Final Render Pipeline

1. Connect the `CONDITIONING` output of your **BatchPromptSchedule** into your advanced sampler (**`SamplerCustomAdvanced`**).
2. Connect your video latent node into the sampler's `latent_image` input.
3. Pipe the sampler's final output into a **`VAE Decode`** node.
4. Replace your static "Save Image" node with a **`Video Combine`** node. Set the container format to `.mp4` and match your target frame rate to **24 fps**.

Once this layout is complete, hitting **Queue Prompt** (`Ctrl + Enter`) will command ComfyUI to render your glitch-transition fashion clip perfectly sequentially. Do you have your three reference outfit photos ready to plug into the graph?
```

### Other

_No response_

No source file, test, or actionable reproduction is identified. First verify that this report describes a ComfyUI problem rather than generated workflow guidance, then replace the placeholder expected and actual behavior with a reproducible workflow and relevant debug logs; done should be a confirmed, reproducible issue with enough information to investigate.

Written by the indexing model from the issue text.