Google Gemini Omni Flash 1.2 is the generally available release of Google's natively multimodal video generation model. It reasons about the scene before rendering and generates native audio such as music, ambience, or dialogue along with the picture, turning plain-language descriptions into smooth, natural short clips. Describe the scene, subject motion, camera work, and sound, and receive an mp4 with a download link. The control panel above the input box lets you pick the resolution (720P by default, or 1080P and 4K, both upscaled from 720P) and the aspect ratio (16:9 landscape or 9:16 portrait). You can upload one PNG or JPG image as a reference (image-to-video), for example a product photo with an orbiting camera move described in the prompt. With two images, describe a transition from the first picture to the second (first and last frame interpolation; the first is the start, the second the end). You can also upload several images and explain which subject or character each one shows (subject references, up to three images per character for up to two characters); the model decides how to use them from your prompt, and at most six reference images are used per run. You can also upload one MP4 video of up to 15 seconds as the source and describe what to change (video editing) or ask the model to continue the scene (video extension); only one video is used per run, and extension is appended to the end of the clip, 3 to 10 seconds at a time and up to 30 seconds in total. It supports continuous conversations (multi-turn editing and extension): after a video is generated, type a follow-up instruction (for example, change the sky to a sunset, slow the pacing, or continue the scene) and the model keeps working on the previous video without re-uploading; you can also upload images in a follow-up turn as references (for example to bring in a new character) and they are used together with the previous video, while uploading a new video starts a fresh generation from that video. Generation runs in the cloud and usually takes one to three minutes; higher resolutions take longer. Great for social clips, product showcases, storyboard previews, and creative animation. For precise results, add film terms such as dolly zoom or slow motion, and describe the lighting, pacing, and background music. Generated videos are MP4 with audio; the resolution is 720P by default or 1080P / 4K (upscaled) from the panel, the aspect ratio is 16:9 landscape or 9:16 portrait, and the clip is 10 seconds by default; describe the length you want in the prompt (officially 3 to 15 seconds, or 3 to 10 seconds for first and last frame interpolation and extension). Prompt example 1: A puppy runs along the waves chasing a frisbee, in slow motion, with the sound of crashing surf and light acoustic guitar music. Prompt example 2: (Upload a JPG product photo of a camping lantern) The camera orbits the product once, with a campsite in morning light behind it, birdsong and gentle ambient sound. Prompt example 3: (Upload an 8-second MP4 street clip) Extend this video: the camera pushes forward to a corner coffee shop entrance, with the sound of passersby chatting. Prompt example 4: (Upload two PNG images: an old street by day and the same street at night) Transition from the first picture to the second as the sky darkens and streetlights come on, rising ambience. Prompt example 5: (Upload two JPG character images: a little penguin and a polar bear) Have the penguin and the polar bear fight over one fish on the ice, cartoon style, playful sound effects. Prompt example 6: (Upload a 6-second MP4 of someone jogging in a park) Change the season to a snowy winter with snow on the ground and crunching footsteps, keeping the runner's movement unchanged. Prompt example 7: (Set the panel to 9:16) A vertical clip: a florist smiles at the camera and says the Mother's Day bouquets are ready, lifting a bunch of carnations, with soft music behind. Prompt example 8: (Type this after the video is generated) Change the background music to soft piano, slow the pacing a little, and end with the camera slowly pulling back. Prompt example 9: Cinematic opening: a foggy harbor at night, the camera does a dolly zoom toward a blinking lighthouse, with a deep cello and a distant ship horn. Prompt example 10: Science animation: a water drop falls from a cloud, flows into a river and evaporates back into the sky, showing the water cycle, with a short English voice-over. Maximum total upload size per request: 512MB (all files and message content combined). Accepted image formats: jpeg, jpg, png. Maximum number of images: 6. Maximum size per image: 32MB. Maximum total image size per message: 32MB (text and history share the same request size limit; uploads close to the limit may still be rejected). Image dimensions: at least 300 pixels per side. Maximum image resolution: 36,000,000 pixels in total. Accepted video formats: mp4. Maximum number of videos: 1. Maximum size per video: 32MB. Video duration: 1 to 15 seconds (total at most 15 seconds). Video dimensions: 64 to 4,096 pixels per side. Video resolution: 4,096 to 9,437,184 pixels in total. Video aspect ratio: 0.25 to 4.0. Video frame rate: 1 to 120 fps. Output content formats: Video Pricing: Medium Video output: one MP4 video with audio; the resolution (720P by default, or 1080P / 4K upscaled) and the aspect ratio (16:9 or 9:16) follow the control panel selection; the length is 10 seconds by default and can be described in the prompt (officially 3 to 15 seconds; first and last frame interpolation and extension 3 to 10 seconds). Continuous conversation: Supported AI can make mistakes. Please verify important information.