Veo3.1 Image To Video (GcsUriImage)

This is the GCS-URI sibling of the tensor-based Veo 3.1 image-to-video node: same generation, but the starting frame is pulled from Google Cloud Storage by URI instead of being passed as an in-graph tensor. If your ComfyUI runs on GKE, or your images live in a bucket as part of an automated pipeline, this node avoids shuttling files around entirely. If you're a local user with images on disk, the tensor version is the friendlier choice - this one's for the cloud-native crowd.

How it works

You give it a gs:// URI and Veo 3.1 uses that image as the first frame, animating per your prompt. The node validates the bucket and object before spending any money - a bad or missing URI produces a clear error. Optional last_frame_gcsuri anchors the end of the clip for interpolation. The finished video downloads to temp and comes out as video_paths (VEO_VIDEO). Like all the Veo 3.1 nodes, it runs the full modern pipeline: model dropdown, resolution choice, native audio, 4–8 second durations.

Inputs that matter

• gcsuri - the required starting-frame image in the cloud.
• last_frame_gcsuri - optional ending frame for guided interpolation.
• model - VEO_3_1, VEO_3_1_FAST, or VEO_3_1_LITE. The lite and fast tiers are cheaper; useful for iteration.
• prompt - the motion description. Empty is tolerated (Veo interprets the image), but a real prompt gets intentional motion.
• output_resolution - 720p (default), 1080p, or 4k. The default is the pack's most common complaint; bump it before a paid run.
• duration_seconds - 4, 6, or 8 only.
• generate_audio - Veo 3.1's native audio, on by default.
• image_format - PNG, JPEG, or MP4.
• compression_quality - lossless requires output_gcs_uri, or the node throws an interactive error demanding a gs:// bucket.

Optional: negative_prompt, seed, output_gcs_uri, gcp_project_id, gcp_region.

Gotchas

Two things carry over from the rest of the Veo 3.1 family. The seed tooltip promises reproducibility with enhance-prompt disabled, but Veo 3.1 doesn't let you disable prompt enhancement - so treat seed as effectively inert. And the 720p default keeps catching people: you're paying for Veo, so set output_resolution to 1080p or 4K unless you're just iterating.

Installing it

Standard pack install:

cd ComfyUI/custom_nodes
git clone https://github.com/GoogleCloudPlatform/comfyui-google-genmedia-custom-nodes
pip install -r comfyui-google-genmedia-custom-nodes/requirements.txt

Or ComfyUI Manager → search comfyui-google-genmedia-custom-nodes. Restart, find it under Google AI/Veo3.1. Set project/region (or rely on GKE workload identity), authenticate, and every video is a billed Vertex AI call. The video_paths output feeds VeoVideoSaveAndPreview for playback or VeoVideoToVHSNode if you want frames back for post-processing.