Generate from text, animate one image, guide a scene with up to ten references, or edit an existing clip with plain-language changes and audio controls.
Powered by Google's preview Gemini Omni video model
Available from the Starter plan. Outputs are 720p at 24 FPS, run 3–10 seconds, and cost 15 credits per second.
Verified examples
Every example below came from the live HummingBytes workflow during launch verification, with no placeholder media.
Start with a written direction and receive a short video with native motion and audio in a wide frame.
Prompt direction
Create a cinematic product scene with controlled camera movement, practical lighting, and subtle ambient sound.
Upload one image, describe the movement, and keep the composition suited to a vertical placement.
Prompt direction
Animate the source with a gentle push-in, natural secondary motion, and a restrained cinematic finish.
Use up to ten images to communicate subjects, materials, styling, and the overall visual direction.
Prompt direction
Combine the supplied references into one coherent landscape scene while preserving their recognizable visual cues.
Edit an eligible MP4 or MOV while HummingBytes normalizes orientation, preserves the source file, and routes the selected audio mode.
Prompt direction
Transform the source into a polished dusk product-stage shot, keep the composition stable, and create a fresh ambient soundtrack.
Where it fits
Use Gemini Omni when the input may change from an idea to a still, a reference set, or an existing clip as the creative direction develops.
Turn a prompt into a short landscape or portrait motion study with native audio.
Add camera and environmental motion to one source image without rebuilding the scene.
Use up to ten images when one reference is not enough to communicate the intended look.
Describe only what should change in an eligible generated or uploaded video.
Gemini Omni Flash is a preview workflow with a deliberately narrow launch contract.
Choose an integer duration from 3 to 10 seconds; longer sequences need separate shots and editing.
The workflow delivers 720p at 24 FPS rather than exposing resolution or frame-rate controls.
Use source audio asks the model to preserve or adapt it, but exact waveform identity is not guaranteed.
Uploaded-video editing and some uploaded-image inputs depend on Google's regional availability.
Model pages are most useful when they connect back into the workflow, use-case, and benchmark pages that help you act on the decision.
Prompt starters
Name the subject, camera behavior, environmental motion, lighting, and audio direction that matter most.
Create a cinematic product launch scene at blue hour. Use a slow camera push, controlled highlights, subtle atmospheric movement, and a quiet original ambient soundtrack. Keep the product readable and the composition practical for a landing-page hero.
Try this promptAnimate the uploaded portrait image with a gentle push-in, natural fabric and hair movement, soft shifting light, and restrained environmental ambience. Preserve the subject and original vertical composition.
Try this promptKeep the subject, timing, composition, and camera movement. Shift the scene from warm dusk to cool moonlight and preserve or naturally adapt the recognizable source soundtrack.
Try this promptReady to make a short video?
Choose the input that best communicates the idea and let one workflow carry it into motion.