Video GenerationActive
google / gemini-omni-flash-r2v
gemini-omni-flash-r2v
Generate 3–10s videos from reference images, or edit an existing clip with one instruction. Built on Google Gemini Omni Flash for coherent motion and style.
Image to VideoVideo to VideoFast Inference
Model ID
gemini-omni-flash-r2v
Provider
google
Updated
1787555480
wiro playground—google/gemini-omni-flash-r2v
Updated 1787555480
Overview
Gemini Omni Flash R2V is a Google video model for reference-guided generation and video editing. It turns one or more reference images plus a written instruction into a short MP4 clip. It can also transform an existing video by applying a simple edit instruction while keeping the scene coherent. This helps you animate a still, keep a look consistent, and iterate on video changes without manual keyframing.
What you can build
- Animate a product photo into a short promo clip with controlled motion and camera direction.
- Generate character shots from reference images for storyboards and social drafts.
- Restyle a real clip into anime, clay, or cinematic looks with a single instruction.
- Make iterative changes to the same generated clip across multiple edit passes.
- Create short vertical or horizontal clips for Shorts, Reels, and TikTok.
Inputs
- Choose whether you want to generate a new video from references, or edit an existing video.
- For reference-guided generation, upload one or more reference images as JPG, JPEG, PNG, or WebP.
- For video editing, upload a single source video as MP4, MOV, WEBM, or M4V.
- Provide a written description of what should happen in the video, or what change to apply.
- When generating from references, select the output length in seconds (supported range is 3–10 seconds).
- When generating from references, select the output framing as landscape (16:9) or portrait (9:16).
Outputs
- An MP4 video file returned as a downloadable result.
- The clip length is 3–10 seconds, depending on the length you requested.
- The model targets 720p output at 24 FPS.
- The output can include an audio track when your instruction asks for sound.
- All generated videos include an invisible SynthID watermark for provenance.
Limitations
- Output duration is limited to 3–10 seconds per generation.
- Only 16:9 and 9:16 aspect ratios are supported for new generations.
- Video extension and frame interpolation are not supported.
- Voice editing is not supported.
- Audio reference uploads are not supported.
- Referencing across multiple videos is not supported and may degrade results.
- Editing uploaded videos is restricted in some regions, including the EEA, Switzerland, and the United Kingdom.
- Low-quality inputs can cause identity drift and artifacts. This includes blurry images, heavy compression, and shaky videos.
Safety & compliance
- Safety filters apply to your text instructions and to generated video.
- Uploading and editing images containing minors is restricted in some regions.
- Uploading and editing media containing certain recognizable people may be restricted.
- Don’t use this model to create disallowed content, impersonation, or deceptive media.
- Treat outputs as synthetic content. Keep the SynthID watermark intact.
Example prompts
Great starting points for gemini-omni-flash-r2v.
The woman from the reference image walks confidently through a sunlit Paris street market, gently turning to look at the stalls, her hair and scarf moving in the breeze, warm morning light, shallow depth of field. Smooth tracking shot following her.Video Generation
The sneaker from the reference image rotates slowly on a glossy studio pedestal, dramatic rim lighting sweeping across it, soft reflections on the floor, subtle dust particles floating in the light beams. Slow orbiting camera, premium product commercial look.Video Generation
The red fox from the reference image trots across a snowy forest clearing, breath visible in the cold air, ears twitching, then pauses and looks toward the camera as light snow drifts down. Slow tracking shot, cinematic wildlife look.Video Generation
The cartoon robot character from the reference image waves cheerfully and bounces on its feet in a bright futuristic lab, small lights blinking on its chest, steam venting playfully from its joints. Gentle push-in, lively animated style.Video Generation
API quick start
Run gemini-omni-flash-r2v with a single API call.
POST https://api.wiro.ai/v1/Run/google/gemini-omni-flash-r2v
{
"prompt": "The woman from the reference image walks …",
"mode": "reference",
"inputImage": "https://your-cdn.com/input.png",
"inputVideo": "https://your-cdn.com/input.mp4"
}