Learning Timeline
Key Insights
Easy Cache Quality Trade-off
Using the 'Easy Cache' node will significantly speed up the generation process, but it comes at the cost of some video quality. Disable it if maximum fidelity is required.
Missing Model Warning
The standard HunyuanVideo 1.5 model card often omits the 'Latent Upsampler' from its requirements list. You must manually download this from the 'split_files' section of the repo, or the upscale workflow will fail.
Step by Step
Downloading and Installing the Latent Upsampler
- Navigate to the official ComfyUI Hugging Face repository.
- Click the 'Files and versions' tab.
- Open the 'split_files' folder.
- Open the 'latent_upscale_models' folder.
- Locate the latent upsampler file (approximately 200MB).
- Click the download arrow next to the filename.
- Move the downloaded file into your ComfyUI installation directory (typically inside 'ComfyUI/models/upscale_models' or the specific folder designated by your workflow).
- Click 'Refresh' on the ComfyUI dashboard to ensure the new model is recognized.
Configuring the 1080p Upscale Workflow
- Locate the Upscaling section/group within the ComfyUI workflow canvas.
- Click the dropdown menu on the main model loader node within this section.
- Select the previously downloaded '1080p upscaler' model.
- Locate the 'Latent Upsampler' node.
- Click the dropdown menu and select the Latent Upsampler model you just installed.
- Locate the 'Easy Cache' node connected to the workflow.
- Configure the Easy Cache settings (enable it to speed up generation, though note this may slightly reduce quality).
- Locate the 'Upscale Method' node.
- Select an algorithm from the list (or retain the default selection).
- Verify the target resolution settings in the Width and Height fields.
- Confirm the 'Frames Per Second' (FPS) value matches your target output.
- Queue the prompt to generate the upscaled video.
