
A WAN 2.2 14B image-to-video workflow. Load one image, write a motion prompt, and get an MP4 video.
Model: DaSiWa WAN 2.2 I2V 14B, version 11. It is a fine-tuned WAN 2.2 model with LoRAs already built in, so you do not need extra physics LoRAs.
How to use:
1. Load your image in the Load Image node.
2. Write your motion prompt in the positive prompt node.
3. Choose the text encoder: FP8 is smaller and uses less memory, BF16 is bigger.
4. Choose a machine and click Run.
Output: an MP4 video (H.264).
My test: a 5-second video took 3 minutes 10 seconds on the Standard machine (24 GB) and cost 38 RH Coins. On my own PC with a 12 GB GPU, this workflow ran out of memory until I optimized it.
Example prompt:
"The woman walks slowly toward the camera through the light rain, holding the clear umbrella steady in her right hand. Raindrops fall through the orange lantern light and splash softly on the wet stones."
For longer videos, increase the frame count. Longer videos take more time and coins.
WAN 2.2 14B Image to Video Workflow (DaSiWa v11)
A WAN 2.2 14B image-to-video workflow. Load one image, write a motion prompt, and get an MP4 video.
Model: DaSiWa WAN 2.2 I2V 14B, version 11. It is a fine-tuned WAN 2.2 model with LoRAs already built in, so you do not need extra physics LoRAs.
How to use:
1. Load your image in the Load Image node.
2. Write your motion prompt in the positive prompt node.
3. Choose the text encoder: FP8 is smaller and uses less memory, BF16 is bigger.
4. Choose a machine and click Run.
Output: an MP4 video (H.264).
My test: a 5-second video took 3 minutes 10 seconds on the Standard machine (24 GB) and cost 38 RH Coins. On my own PC with a 12 GB GPU, this workflow ran out of memory until I optimized it.
Example prompt:
"The woman walks slowly toward the camera through the light rain, holding the clear umbrella steady in her right hand. Raindrops fall through the orange lantern light and splash softly on the wet stones."
For longer videos, increase the frame count. Longer videos take more time and coins.
A WAN 2.2 14B image-to-video workflow. Load one image, write a motion prompt, and get an MP4 video.
Model: DaSiWa WAN 2.2 I2V 14B, version 11. It is a fine-tuned WAN 2.2 model with LoRAs already built in, so you do not need extra physics LoRAs.
How to use:
1. Load your image in the Load Image node.
2. Write your motion prompt in the positive prompt node.
3. Choose the text encoder: FP8 is smaller and uses less memory, BF16 is bigger.
4. Choose a machine and click Run.
Output: an MP4 video (H.264).
My test: a 5-second video took 3 minutes 10 seconds on the Standard machine (24 GB) and cost 38 RH Coins. On my own PC with a 12 GB GPU, this workflow ran out of memory until I optimized it.
Example prompt:
"The woman walks slowly toward the camera through the light rain, holding the clear umbrella steady in her right hand. Raindrops fall through the orange lantern light and splash softly on the wet stones."
For longer videos, increase the frame count. Longer videos take more time and coins.

