All of these nodes need a Google Cloud project and region as input and uses the Vertex AI APIs to access Google GenMedia models. If you are running ComfyUI on GKE as described in ComfyUI on GKE guide, the Vertex AI APIs will be authenticated with GKE workload identity and no additional action is required form your end.

If you are running your own(local) ComfyUI installation, you will need to install the custom nodes and authenticate to Google Cloud as an user or a Service Account. See installation section for details.

Pricing

All the custom nodes provided with this repo use Vertex AI APIs to generate content and hence, Vertex AI pricing is applied to each API call. See Vertex AI Pricing for the model that you want to use.

Note : If you store the content in Google Cloud Storage using these custom nodes, refer to Google Cloud Storage pricing.

Installing the Google Generative Media custom nodes

ComfyUI on GKE

If you want to run ComfyUI on GKE and use these custom nodes, follow the ComfyUI on GKE guide that will spin up ComfyUI on GKE with these custom nodes installed and ready to use.

ComfyUI running locally

If you are running ComfyUI locally or at a place other than GCP and want to use these custom nodes, you need to perform the following steps to install them:

Download these custom nodes in one of the following ways:
- Using the ComfyUI Manager
  - Go to ComfyUI Manager --> Custom Node Manager --> search for comfyui-google-genmedia-custom-nodes. Click Install.
- Manual installation using git
  - Open the terminal on the machine that is running ComfyUI and change to the directory where ComfyUI is running from.
```
cd <PATH_TO_COMFYUI_INSTALLATION>

E.g if your ComfyUI is running from "/home/user/username/ComfyUI", replace <PATH_TO_COMFYUI_INSTALLATION> with "/home/user/username/"
```
  - Change to the custom_nodes folder.
```
 cd ComfyUI/custom_nodes
```
  - Clone this repository in the custom_nodes folder:
```
git clone https://github.com/GoogleCloudPlatform/comfyui-google-genmedia-custom-nodes
```
  - Install python packages:
    1. If you are running ComfyUI in a python virtual environment, activate it before installing the dependencies.
```
  source <PATH_TO_PYTHON_VIRTUAL_ENVIRONMENT>/bin/activate
```
    2. Install python packages
```
pip install -r comfyui-google-genmedia-custom-nodes/requirements.txt
```
Now that the custom nodes are installed, set up authentication with Google Cloud. This is needed because the custom nodes will use Vertex AI APIs in the given Google Cloud project and region. Choose one of the following authentication methods:
- Authenticate local ComfyUI with Google Cloud user
- Authenticate local ComfyUI with Google Cloud Service Accountt
Re-start ComfyUI via ComfyUI manager or manually. Verify there are no failures in the logs and the log says that google-genmedia custom node has been loaded successfully, similar to the following message:
```
Import times for custom nodes:
   .....
   .....
   .....
   2.8 seconds: /XXX/XXXX/ComfyUI/custom_nodes/comfyui-google-genmedia-custom-nodes
```
You are all set! Start using the custom nodes. Right click on ComfyUI, Click the 'Add node' menu and you should see a category named Google AI with custom nodes for Google’s genmedia models as shown in the image below. You will need to pass a GCP project id and GCP region as input parameter to every node if you are using the custom nodes from your local ComfyUI installation.
Optional step : If you are not familiar with Google's GenMedia custom nodes, we provide some sample workflows with this repo that shows how to use the nodes. You can download and use the sample workflows by following the steps in Sample Workflows section. For full set of features and input parameters to each node, see Node features section for more details.

Authenticate local ComfyUI with Google Cloud user

On the machine where you are running ComfyUI, install gcloud cli using the documentation https://docs.cloud.google.com/sdk/docs/install
Authenticate with the following command. Enter your Google Cloud credentials when prompted:
```
  gcloud auth application-default login
```

Authenticate local ComfyUI with Google Cloud Service Account

On the machine where you are running ComfyUI, install gcloud cli using the documentation https://docs.cloud.google.com/sdk/docs/install
Authenticate with the following command. Enter your Google Cloud credentials when prompted:
```
 gcloud auth login
```

Set shell variables

export PROJECT_ID="<YOUR_PROJECT_ID>"
export SERVICE_ACCOUNT_ID="<YOUR_SERVICE_ACCOUNT_ID>"
export USER_EMAIL=`gcloud auth list --filter=status:ACTIVE --format="value(account)"`

Replace <YOUR_PROJECT_ID> with the GCP project id you plan to use
Replace <YOUR_SERVICE_ACCOUNT_ID> with the name of the service account you will use to access Google Cloud with ComfyUI. It will be created in the next step.

Set Google Cloud project.
```
gcloud config set project $PROJECT_ID
```

Create a new service account:

gcloud iam service-accounts create $SERVICE_ACCOUNT_ID \
--display-name="ComfyUI Service Account" \
--description="Service Account to use ComfyUI"

Grant permissions to the service account:

gcloud projects add-iam-policy-binding $PROJECT_ID \
--member="serviceAccount:${SERVICE_ACCOUNT_ID}@${PROJECT_ID}.iam.gserviceaccount.com" \
--role="roles/storage.objectUser" --condition=None

gcloud projects add-iam-policy-binding $PROJECT_ID \
--member="serviceAccount:${SERVICE_ACCOUNT_ID}@${PROJECT_ID}.iam.gserviceaccount.com" \
--role="roles/aiplatform.user" --condition=None

Grant the Service Account Token Creator role to your user on the service account. This role lets your user account to impersonate the Service Account created in previous steps:

gcloud iam service-accounts add-iam-policy-binding \
${SERVICE_ACCOUNT_ID}@${PROJECT_ID}.iam.gserviceaccount.com \
--member="user:${USER_EMAIL}" \
--role="roles/iam.serviceAccountTokenCreator"

Now, set yourself to impersonate the service account

gcloud auth application-default login --impersonate-service-account=${SERVICE_ACCOUNT_ID}@${PROJECT_ID}.iam.gserviceaccount.com

Node features

Gemini 2.5 Custom Node

Category: Google AI/Gemini

This node integrates the Gemini 2.5 into ComfyUI, supporting text generation only with multimodal inputs (image, video, audio).

Required Inputs

Optional Inputs

Outputs

Sample

Sample workflow using Gemini 2.5 node

Gemini 3.1 Custom Node

Category: Google AI/Gemini

This node integrates the Gemini 3.1 into ComfyUI, supporting text generation with multimodal inputs (image, video, audio).

Required Inputs

Optional Inputs

Outputs

Sample

Sample workflow using Gemini 3.1 node

Gemini 3.5 Custom Node

Category: Google AI/Gemini

This node integrates the Gemini 3.5 into ComfyUI, supporting text generation with multimodal inputs (image, video, audio).

Required Inputs

Optional Inputs

Outputs

Sample

Sample workflow using Gemini 3.5 node

Veo2 Custom Nodes

1. Veo2 Text To Video

Category: Google AI/Veo2

Generates videos purely from text prompts using the Google Veo 2.0.

Required Inputs

Optional Inputs

Outputs

Sample

2. Veo2 Image To Video (GcsUriImage)

Category: Google AI/Veo2

Generates videos using an image stored on Google Cloud Storage (GCS) as the starting frame.

Required Inputs

Optional Inputs

Outputs

3. Veo2 Image To Video (Tensor)

Category: Google AI/Veo2

Generates videos using a ComfyUI image tensor (passed from another node) as the starting frame.

Required Inputs

Optional Inputs

Outputs

Veo 3.1 Custom Nodes

Veo3.1 Text to Video

Category: Google AI/Veo3.1

Generates videos purely from text prompts using the Google Veo 3.1.

Required Inputs

Optional Inputs

Outputs

Sample

Sample workflow using Veo3 Text to Video node

2. Veo3.1 Image To Video (GcsUriImage)

Category: Google AI/Veo3.1

Generates videos using an image stored on Google Cloud Storage (GCS) as the starting frame.

Required Inputs

Optional Inputs

Outputs

Sample workflow using Veo3 Image to Video node

3. Veo3.1 Image To Video (Tensor)

Category: Google AI/Veo3.1

Generates videos using a ComfyUI image tensor (passed from another node) as the starting frame.

Required Inputs

Optional Inputs

Outputs

Sample

Sample workflow using Veo3 Image to Video node

4. Veo3.1 Reference To Video

Category: Google AI/Veo3.1

Generates videos using multiple reference images to guide the style and content.

Required Inputs

Optional Inputs

Outputs

Sample

Sample workflow using Veo3 Reference to Video node

Gemini 2.5 Flash Image (Nano Banana 🍌)

Category: Google AI/GeminiFlashImage

This node integrates the Gemini 2.5 Flash Image model, designed for high-speed image generation and editing tasks.

Required Inputs

Optional Inputs

Outputs

Sample

Sample workflow using Gemini 2.5 Flash node

Gemini 3.1 Flash Image (🍌)

Category: Google AI/GeminiFlashImage

This node integrates the Gemini 3.1 Flash Image model, designed for high-speed image generation and editing tasks.

Required Inputs

Optional Inputs

Outputs

Sample

Sample workflow using Gemini 3.1 Flash node

Gemini 3 Pro Image (Nano Banana Pro 🍌)

Category: Google AI/GeminiProImage

This node integrates the Gemini 3 Pro Image model, designed for high-speed image generation and editing tasks.

Required Inputs

Optional Inputs

Outputs

Sample

Sample workflow using Nano Banana Pro node

Virtual Try-On

Category: Google AI/Virtual try-on

This node leverages Google's Virtual Try-On model. It takes an image of a person and an image of a garment (product) and generates a new image of the person wearing that garment.

Required Inputs

Outputs

Sample

Sample workflow using Virtual-try-on node

Lyria 2 Text To Music

Category: Google AI/Lyria2

This node integrates with Lyria2 to generate music tracks based on text prompts.

Required Inputs

Optional Inputs

| Parameter | Type | Default | Description | | :--- | :--- | :--- | :--- | | negative_prompt | STRING | "" | Text description of musical elements to avoid (e.g., "drums", "vocals"). |

Outputs

Video to VHS

Category: Google AI/Utils

This node takes the video output from a Veo generation node, extracts frames (up to 120), and converts them into a batch of images (Tensor). This is useful for processing video frames with other ComfyUI image-to-image or VAE nodes (often referred to as "VHS" workflows in ComfyUI).

Required Inputs

| Parameter | Type | Default | Description | | :--- | :--- | :--- | :--- | | video_paths | VEO_VIDEO | N/A | The internal list of video file paths passed from a Veo generation node. |

Outputs

Preview/Save video

Category: Google AI/Utils

This node handles the display and persistence of generated videos. It allows you to preview the videos generated by Google's models in the ComfyUI interface and optionally save the file permanently to the output directory.

Required Inputs

Outputs

This node does not produce data outputs to connect to other nodes. It outputs the video player UI widget.

Sample Workflows

Here is a GIF showing the various workflows in action, Workflow demo

Download and use sample workflows provided with this repo:

ComfyUI on GKE - If you are running ComfyUI via ComfyUI on GKE guide, the sample workflows will be provided with the installation. Click onWofkflows on ComfyUI screen and you will see the list of all the sample workflows. Click on any workflow and run it as shown in the GIF above.

ComfyUI running locally - If you are running ComfyUI locally or at a place other than GCP, you need to perform the following steps to download the sample workflows. These steps assume, you have already downloaded this repo by following the instructions in ComfyUI running locally

Open the terminal on the machine that is running ComfyUI and change to the directory where ComfyUI is installed.

cd <PATH_TO_COMFYUI_INSTALLATION>

E.g if your ComfyUI is running from "/home/user/username/ComfyUI", replace <PATH_TO_COMFYUI_INSTALLATION> with "/home/user/username/"

Change to the custom_nodes folder.
```
cd ComfyUI/custom_nodes
```

Copy the workflows by running the following command:

mv comfyui-google-genmedia-custom-nodes/sample-workflows/*json ../user/default/workflows/

Install dependencies for the workflows to function:

mv comfyui-google-genmedia-custom-nodes/sample-workflows/input-images/* ../input/

Open ComfyUI manager,click Custom Nodes Manager and search for ComfyUI-VideoHelperSuite and install the latest version. Then search for ComfyUI-Custom-Scripts and install the latets version. These are required as we use these custom nodes in the example workflows.
Restart ComfyUI via ComfyUI manager or manually. Check in the left side menu under workflows, the new .json workflow files should be present now. You can click on the files to load the workflows and then RUN them.
Make sure to pass the gcp_project_id and gcp_region to all the custom nodes in the workflows.

Note that some of these workflows require an input parameter output_gcs_uri. This is required in scenarios such as when you are generating a lossless video using veo3 and the APIs can not return a huge video file so it expects an output bucket to be provided as input where it saves the output video. The custom node will throw an interactive error saying output_gcs_uri must be provided. If you encounter such situation, provide the bucket in the input parameter output_gcs_uri in the format gs://YOUR_BUCKET_NAME to the custom nodes. If you do not have a GCS bucket, create it in your Google Cloud project first.