This is an Image to Video workflow using WAN2.2 A14B fp8

This workflow enables you to quickly use multiple LoRAs within a single workflow by duplicating and saving bundles of LoRAs. It serves as a template for customizing the LoRA loader settings to your liking and creating your own workflow. It also includes several sample LoRA bundles

TIME COST

The generation time is approximately 2 to 3 minutes by default.

When using End image, generation time takes approximately 4 to 6 minutes.

Generation times may vary depending on the settings. Additionally, generation times may change during peak service periods. Please consider this time as an estimate only.

HOW TO USE

(1).LoRA

LoRA is the most crucial part of this workflow. The explanation will be a bit lengthy, but please read through to the end.

This workflow contains multiple LoRA groups, each bundling several LoRA loaders.the purpose of this workflow is to enable switching between multiple LoRA bundles by switching these LoRA groups.

The LoRA group contains 3 LoRA Stuck nodes. Each Stuck node can load 5 LoRAs, each LoRA Stuck node is designated for either Share, High noise, or Low noise. Since Wan 2.2 generates videos using two models (High noise and Low noise) and two samplers, the LoRA is also split into High noise and Low noise versions. Load the High noise LoRA into the node labeled High, and load the Low noise LoRA into the node labeled Low. Share is for LoRAs not divided into High/Low, such as those used by Wan 2.1 or AIO.

While the three Stuck nodes provide a maximum of 15 slots, please note that only up to 10 LoRAs can be loaded at once.

There are 4 empty LoRA groups, which you can freely customize and save LoRA presets to your workflow.

1.1. Load LoRA

When loading LoRA, clicking the section labeled “lora_<number>” will open a window for selecting LoRA.

Note : that the first node in the High/Low section always has Lightx2 LoRA loaded. This LoRA accelerates generation and is related to the sampler settings. If you do not understand this LoRA, please do not it.

Unlike other Stuck nodes, there is no switch to individually turn slots ON/OFF. All loaded LoRAs will be enabled.

If you wish to disable a LoRA once loaded, set its Strength to 0.

Strength sets the intensity of the LoRA. Typically, it is set to around 0.8 to 1, but if specific requirements exist for each LoRA, follow those instead.

1.2. Switch LoRA Groups

To enable the LoRA group, connect the “lora” port of the High/Low Stuck nodes to the Set_LoRA_High and Set_LoRA_Low nodes, respectively. Connect High to High and Low to Low.

Remember to bypass unused groups, There are several ways to bypass unnecessary groups.

The simplest is to press the switch on the red node labeled “Fast Groups Bypasser (rgthree)” within the group. Press the white button labeled yes/no.

If the entire node turns purple, it means it has been bypassed.

Other bypass methods, select the group and press Ctrl + B. or Right-click and select Bypass.

1.3. Samples Groups

As samples groups, several LoRA bundles are included.

The sample group comes preloaded with LoRA and includes a note containing the LoRA page URL, trigger word, and sample prompt.

Simply connect the group and copy-paste the sample prompt to start using it immediately.

Many groups are arranged across a wide area, and you may struggle to connect pipes to the groups at the edges and the Set_LoRA node.

You can move entire groups along the workflow by dragging their headers. This allows you to rearrange High-frequency groups closer.

1.4. Bypasser Setting

If you duplicate a LoRA group and create a new group, fine-tuning is required for the Fast Groups Bypasser.

Right-click the newly created Bypasser and open its Properties panel.

Enter the name of the newly created LoRA group in the Match title field. This will make the Bypasser correspond to that group.

To name a group, double-click the title bar.

That concludes the explanation of LoRA. Thank you for your hard work.

(2).Prompt

Enter the prompt.

(3).Load Image

Upload an image to be the first frame of your video.

(4).Config

Set the video length (number of frames) and resolution (long side).

Videos are generated at 16 fps by default, so the number of frames is calculated as a multiple of 16. The default setting of 81 frames is approximately 5 times 16, or 5 seconds.

Set the resolution to the number of pixels on the long side. The resolution on the short side will be set to the original resolution to match the long side.

However, this value must be a multiple of 16.

If it is not a multiple of 16, you may see the error "RuntimeError: The size of tensor a..."

In this case, adjusting the aspect ratio of the ImageScaleByAspectRatio (FST) node to the scale factor closest to the loaded image may resolve the issue.

(5).End Image

Loads the image that will be the last frame of the video.

This function is off by default (Bypassed). To use it, press the switch on this node.

Once all the settings are complete, click the "Generate" button to begin generation.

(6).SAVE & LOAD

Once you've configured your preferred settings, you can save that state for later use.

To save, press the Save button.

Use Save As to create a new save, and Save to save the currently open workflow.

Next time you want to use a workflow, you can always access the saved workflows from “My Workflows.”

Once you've customized your workflow and created something you're proud of, go ahead and publish it.

That concludes the explanation. Thank you for reading this far. Enjoy!

Wan2.2 A14B fp8 Multi-LoRa Group Template

Rating & Review