Nodes/ComfyUI-Google-AI-Studio/Google AI Studio Omni Video
ComfyUI Node

Google AI Studio Omni Video

Generate or edit videos with Gemini Omni Flash. Returns a 24 fps IMAGE batch, the native audio track, and the interaction ID for follow-up edits.

By BuffMcBigHuge·Created about a year ago·Updated 4 days ago· 11
Google AI Studio Omni Video
  • first_frame
  • last_frame
  • reference_images
  • frames
  • audio
  • interaction_id
◄promptA continuous, unbroken handheld shot of a fluffy tabby cat sitting on a sunny windowsill, looking out into a leafy garden. The cat's tail twitches slowly. Sound design: gentle breeze, distant bird chirps. No dialogue.►
◄api_key►
◄modelgemini-omni-1.2-flash►
◄aspect_ratio16:9►
◄resolution720p►
◄duration10►
◄taskauto►
◄previous_interaction_id►
◄deliveryinline►
◄frame_step1►
◄max_side1280►
◄model_override►
CategoryGoogle AI Studio

Inputs (15)

NameTypeDefaultDescription
promptSTRINGA continuous, unbroken handheld shot of a fluffy tabby cat sitting on a sunny windowsill, looking out into a leafy garden. The cat's tail twitches slowly. Sound design: gentle breeze, distant bird chirps. No dialogue.—
api_keySTRINGGoogle AI Studio API key. Leave empty to use GOOGLE_API_KEY / GEMINI_API_KEY.
modelCOMBOgemini-omni-1.2-flash2 options: gemini-omni-1.2-flash, gemini-omni-1.1-flash
first_frameoptIMAGEOptional starting image. With last_frame, creates an interpolation.
last_frameoptIMAGEOptional ending image for first/last-frame interpolation (duration must be 3-10s).
reference_imagesoptIMAGEOptional reference image batch (up to 6) for character/object consistency.
aspect_ratiooptCOMBO16:92 options: 16:9, 9:16
resolutionoptCOMBO720p1080p and 4k are upscaled by the API.
durationoptINT103–15Video length in seconds. First/last frame interpolation supports 3-10s.
taskoptCOMBOautoUsually leave on 'auto'. 'edit'/'extend' require a previous_interaction_id.
previous_interaction_idoptSTRINGInteraction ID from an earlier Omni run, for conversational editing or extension.
deliveryoptCOMBOinlineUse 'uri' for videos larger than ~4 MB (720p and above).
frame_stepoptINT11–10Keep every Nth frame to reduce memory. 1 keeps all frames (24 fps); N>1 returns 24/N fps, so set VHS frame_rate to 24/N.
max_sideoptINT12800–8192Longest side of the returned frames in pixels. 0 keeps the original resolution.
model_overrideoptSTRINGOptional model ID that overrides the dropdown (for brand-new models).

Outputs (3)

NameTypeDescription
framesIMAGE—
audioAUDIO—
interaction_idSTRING—