Create clips up to 15 seconds from text or images, then edit or extend your scenes with Gemini Omni 1.2 Flash. Direct the motion and sound, and check the credit cost before generating.
Text to video · First and last frames · Visual references · Video editing · Scene extension
Currently creating with Gemini Omni 1.2 Flash
VIDEO STUDIO
Your video will appear here after you generate.
Track a loose orange at pavement level as it escapes a fallen canvas shopping bag on a lively European market street. Follow its rolling path between café furniture, across a shallow puddle, and beneath a vendor’s cart. A passing shopper stops the fruit, lifts it, and hands it to a smiling woman in a light jacket. Rise slowly with the handoff to reveal the market around them. Late-afternoon backlight catches the orange peel and wet paving stones. Keep the movement fluid, the textures lifelike, and the background softly out of focus. Include footsteps, gentle street chatter, and the sound of the fruit rolling.
Google's video model turns written directions and visual references into clips with sound.
Gemini Omni 1.2 Flash is a Google AI model for generating and editing video. You can describe a new scene or request changes to an existing clip.
Gemini Omni 1.2 Flash is a Google video model that accepts written directions and visual inputs. On JXP, its five modes cover a new scene from text, an animated image, a scene guided by references, a focused video edit, and a continuation of an existing shot.
Choose the mode before preparing your files. A starting image fixes the opening of a shot; a reference image guides its appearance. A reference video suggests movement, while a source video is the clip you are actually editing or extending. Keeping those roles separate makes your request easier to control.
Build the first shot, keep your references in view, and make focused changes as your idea develops.
Keep the controls close to the creative decision. Start with the material you have, then choose the length, framing, and output resolution for that particular shot.
Describe one subject, one main action, and a camera position. Text to Video accepts no media files and supports 3–15 seconds. Add sound direction in the same prompt, such as a quiet workshop, footsteps on gravel, or a short spoken line.
Image to Video uses one starting image, with an optional ending image. One image supports 3–15 seconds; a first-and-last-frame pair supports 3–10 seconds. Choose frames with compatible composition and explain how the subject moves between them. A dramatic change in viewpoint may make continuity harder.
Reference mode accepts up to 10 images and 3 short videos through this integration, with a 3–10 second output. Use images for a subject, location, or material and short videos for motion. The uploader accepts reference clips up to 3 seconds each. Frame images and reference media belong to separate modes.
Video Edit takes one source clip up to 15 seconds and optional reference images. Ask for a specific change, such as warmer lighting or a new background, while naming the elements to preserve. The tool reads the source duration to calculate its credit cost.
Extend Video adds 3–10 seconds to a source clip, within a 30-second total. Describe the next action from the existing endpoint: a door opens, the subject turns, or the camera reveals the next part of a room. The extension length is the number of new seconds, not the target length of the complete clip.
Output choices are 360p, 720p, 1080p, and 4K. Text, image, and reference generation offer 16:9 landscape or 9:16 portrait; image and reference modes also let you follow the input framing. Editing and extension keep the source framing. Export in MP4, WebM, or MOV.
The input and duration controls change with the task you select.
Use this quick reference before uploading. First and last frames cannot be mixed with reference images or reference videos in the same request.
| Mode | Inputs | Duration |
|---|---|---|
| Text to Video | Prompt only | 3–15 seconds |
| Image to Video | 1 first frame; optional last frame | 3–15 seconds; 3–10 with two frames |
| References | Up to 10 images and 3 videos | 3–10 seconds |
| Video Edit | 1 source video; up to 10 optional images | Source length, up to 15 seconds |
| Extend Video | 1 source video | Add 3–10 seconds; 30 seconds total |
The workspace checks file type, size, and readable media metadata before uploading. Use JPG, PNG, or WebP images up to 10 MB each, and MP4, WebM, or MOV videos up to 100 MB. Keep a clear subject and avoid supplying several contradictory views of the same object.
Plan product clips, social ads, short scenes, and visual variations around the material you already have.
Give each video one clear job. A product shot can show an item in context; a short scene can test an idea before you develop a longer sequence.
Start with a clean product image and one controlled camera move: a slow push toward a bottle, a gentle orbit around a lamp, or a close-up of a fabric texture. Describe the surface, lighting, and background. Review logos, labels, and edges frame by frame before using the clip in a campaign.
Choose 9:16 before generating a new scene. Place the action near the middle of the composition and leave space where captions will go later. A short, readable action often works better than packing several cuts, locations, and messages into one clip.
Supply an opening image and an ending image in Image to Video. Describe the physical transition between them, including what moves first and how the camera behaves. Matching framing and subject placement gives the transition a more coherent starting point.
Use References when a character, object, location, or visual treatment matters more than an exact opening frame. Give each reference a purpose in the prompt. Compare the result against the supplied details; references guide generation but do not guarantee a pixel-for-pixel match.
A practical option for creators, marketing teams, and sellers making short videos from an idea or existing assets.
Start with the material you already have and the result you need. A clear brief makes it easier to choose between generating a scene and editing a clip.
Turn a scene description into a draft you can inspect. Use the result to decide which framing, action, or mood to develop next.
Try a new setting or visual direction around the same product. Review each version against the campaign brief rather than assuming every variation is ready to publish.
Give a still product image a setting and movement. Check important details, especially packaging and labels, before using the clip on a listing or in an ad.
Prepare your input, describe the shot, generate a draft, and refine the details you want to change.
Create a new clip or work from an existing video without leaving this page. The workspace shows which inputs and output settings apply to your selected mode.
Select text, image, references, edit, or extend. Add the files your mode needs and describe the subject, action, camera movement, and sound.
Choose resolution, duration, and framing. Review the credit cost before generating. Eligible free vouchers apply to 360p and 5 seconds.
Generate your video, review the finished scene, and download the result. Use Resume checking if status updates stop, or start a focused edit to refine your clip.
Choose a credit plan that fits your video projects.
Pay once and use credits anytime - they never expire
Get fresh credits every month with automatic renewal. Subscription plans offer 20% more credits compared to one-time purchases.
Practical answers about input files, output controls, credits, and generation.
These answers describe the controls available in this JXP workspace. They distinguish the five modes so you can prepare the right files and budget for the result.
Text to Video and single-image generation support 3–15 seconds on JXP. Two-frame and reference generation support 3–10 seconds. Editing accepts a source video up to 15 seconds; extension adds 3–10 seconds within a 30-second total. Pick the mode first so the duration control matches the task.
Eligible free vouchers cover a 360p request lasting exactly 5 seconds. Otherwise, generation uses credits at 1 per second for 360p, 2 for 720p, 3 for 1080p, or 6 for 4K. Your account must have a voucher or enough credits for the selected settings. Each new generation or revision is a separate request.
Use first and last frames in Image to Video, and use reference images or short reference videos in References mode. These input types cannot be mixed in one request through this integration. For a two-frame shot, upload the beginning image first and the ending image second; then explain the motion between them.
References mode accepts up to 10 images and 3 video clips through this integration. The uploader accepts reference videos up to 3 seconds each, and reference-mode output is 3–10 seconds. Video Edit separately allows one source video plus up to 10 optional images. Limits can differ between providers, so use the controls on this page when preparing your files.
Editing changes an existing clip, while extension adds action after its ending. Editing is billed using the uploaded source duration rounded up to a whole second; extension is billed using the additional seconds you choose. Both use the source framing. Reference videos are guidance for a new scene, not substitutes for the source video in these modes.
Yes. Text, image, and reference generation offer 16:9 landscape or 9:16 portrait, with 360p, 720p, 1080p, and 4K output options. Image and reference modes can also follow the input framing. Editing and extension do not expose a new aspect-ratio control. Preview a lower-resolution version before spending credits on a larger output if you are still deciding on the scene.
This workspace does not accept uploaded audio. Describe desired speech, ambience, music, or sound effects in your video prompt. Keep sound directions consistent with the scene, and review the generated audio along with the visuals before downloading your final version.
Use Resume checking to query the existing task again. If a submission was interrupted before its task ID was returned, check Generation history before starting another request. Refreshing the status does not create another video; submitting a new request may use more credits.
Start with one scene and a clear direction.
Try the example prompt, animate a product image, or upload a clip to refine. Choose the mode that fits your material and check the credit cost before generating.