Wan: Animate character with InfiniteTalk

Step-by-step

1. Navigate to the 12.1 Infinite Talk 14B single speaker section
2. Select text prompt only and upload your character image
3. Select one speaker only and upload the audio file
4. Unselect ignoring background music
5. Enter a short text prompt and set the aspect ratio to 9 by 16
6. Set the frame count based on audio length (e.g., 350 frames for 14 seconds)
7. Configure inference steps to 10 and set audio guidance to 7
8. Go to the LoRAs tab, select the Fusion X LoRA, and choose TCH
9. Click generate

Options

• Use a curl command to download and place the Fusion X LoRA in the correct folder
• Adjust the seed, guidance, and shift values

Watch out for

• You must download the Fusion X LoRA to use lower inference steps effectively

Tips

• Use the Fusion X LoRA to get the tool to work faster and better
• Calculate frames by multiplying the frame rate (25) by the audio duration in seconds
• An audio guidance of 7 is recommended for good results
• Don't bother with the 'ignoring background music' setting if your audio doesn't have any

Highlights

“it puts it in the right folder and is a quick way to get it set up”