Configure and generate Text-to-Video with HunyuanVideo 1.5 in ComfyUI | Alpha | PandaiTech

Configure and generate Text-to-Video with HunyuanVideo 1.5 in ComfyUI

How to select models, set prompts, configure resolution/steps, and run the text-to-video workflow in ComfyUI.

Learning Timeline
Key Insights

Easy Cache Trade-off

Enabling the 'Easy Cache' node can significantly speed up the workflow, but it comes at the cost of reduced video quality.

Frame Calculation Formula

To determine the value for the 'Length' input: Target Duration (Seconds) x 24 = Frame Count.

Bypass Shortcut

Use 'Ctrl + B' (or Cmd + B on Mac) to quickly toggle nodes On/Off (Bypass/Unbypass) when switching between different decoding methods.
Prompts

Example Action Prompt

Target: ComfyUI Positive Prompt Node
a giant cat wearing aviator goggles races across a desert with a child clinging on its back. The cat runs so fast that it leaves a trail of fire.
Step by Step

Configuring Model Loaders and Prompts

  1. Press the 'R' key on your keyboard to refresh the ComfyUI models list.
  2. Locate the main loader node (labeled 'Step 1' in the video workflow).
  3. Click the first dropdown menu and select the main HunyuanVideo model file.
  4. Click the second dropdown menu and select the 'coin 2.5' model.
  5. Click the third dropdown menu and select the 'bite 5 small' model.
  6. Click the VAE dropdown menu and select the appropriate VAE file.
  7. Navigate to the text input box labeled for the positive prompt and type your desired video description.
  8. Navigate to the negative prompt text box and enter elements to exclude from the video.

Setting Resolution and Generating

  1. Locate the resolution and frame settings node.
  2. Verify the frame rate is set to 24 fps.
  3. Calculate the 'length' value (Total Frames) by multiplying desired seconds by 24 (e.g., set to 120 for a 5-second video).
  4. Set the 'batch_size' value (e.g., 1 for a single video generation).
  5. Click the 'Queue Prompt' or 'Run' button in the ComfyUI menu to start generation.

Optimizing Speed with VAE Decoding

  1. Locate the 'VAE Decode High' node if generation is too slow.
  2. Select the 'VAE Decode High' node and press 'Ctrl + B' to unbypass (activate) it.
  3. Locate the original 'VAE Decode' node (often the Tiled version).
  4. Select the original node and press 'Ctrl + B' to bypass (deactivate) it to ensure only one decoder runs.

More from Generate Commercial & Cinematic AI Video

View All