Gemini Omni Flash Reference to Video
Generate 3-10 second videos from up to 10 reference images with Google's Gemini Omni Flash. Keep subjects, products, and style consistent across shots while the model composes a new scene — with dialogue, music, and sound effects rendered natively and directed through the prompt.
69% less than the priciest provider (5 compared). Same model, same output.
Up to 10 reference images for subjects, products, style or composition.
Duration
Length of the generated video in seconds (3-10). Billed per second.
Resolution
720p is native; 1080p and 4k upscale the 720p render (1080p at no extra cost).
Aspect Ratio
Aspect ratio of the generated video.
69% less than the priciest provider (5 compared). Same model, same output.
Your result renders here.
Use Gemini Omni Flash Reference to Video via the API
Generate with Gemini Omni Flash Reference to Video from your own code. Async by design: submit a generation, get a run id back instantly, poll until it completes. Billed in credits at the best available provider price — from $0.05 per second.
1 — Start a generation
curl -X POST https://app.scenetra.com/api/v1/generate/gemini-omni-flash-ref2v \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{"prompt":"The person from the first image presents the product from the second in a bright studio, gentle music","image_urls":["https://example.com/person.jpg","https://example.com/product.jpg"],"duration":"8"}'Response (202)
{
"runId": "run_abc123",
"status": "PENDING",
"model": "gemini-omni-flash-ref2v",
"estimatedCredits": 400,
"statusUrl": "/api/v1/runs/run_abc123"
}2 — Poll until COMPLETED (every 2-5 seconds)
curl https://app.scenetra.com/api/v1/runs/run_abc123 \
-H "Authorization: Bearer YOUR_API_KEY"Response
{
"runId": "run_abc123",
"status": "COMPLETED",
"progress": { "completed": 1, "total": 1 },
"outputs": [
{ "type": "video", "url": "https://images.scenetra.com/..." }
],
"error": null,
"estimatedCredits": 400
}Request fields
| Field | Type | Description | Default |
|---|---|---|---|
| prompt* | string | Describe the result and how the reference images should be used (subjects, style, composition). Audio is rendered natively. | — |
| image_urls | string[] (https URLs) | Up to 10 reference images for subjects, products, style or composition. | — |
| duration | string | Length of the generated video in seconds (3-10). Billed per second.3 · 4 · 5 · 6 · 7 · 8 · 9 · 10 | 8 |
| resolution | string | 720p is native; 1080p and 4k upscale the 720p render (1080p at no extra cost).720p · 1080p · 4k | 720p |
| aspect_ratio | string | Aspect ratio of the generated video.16:9 · 9:16 | 16:9 |
| callbackUrl | string (https URL) | Optional. When the run finishes (completed or failed), we POST the final run status to this URL — same shape as the poll response, plus model, params, and chargedCredits. Best-effort; polling remains the source of truth. | |
Unknown fields are rejected with a helpful 400. Media fields take https URLs. There is no provider parameter — every run is routed to the cheapest capable provider automatically.
Good to know
Billing
Runs bill your account credits at provider cost with 0% markup. The submit response includes the credit estimate; failed generations are refunded automatically. Add ?dryRun=1 to get the estimate without running anything.
Errors
Errors are always { "error": { "code", "message" } } with stable codes: unauthorized, model_not_available, invalid_parameter, insufficient_credits, rate_limited. A failed run is not an HTTP error — poll responses report status: "FAILED" with the reason.
What makes Gemini Omni Flash Reference to Video different
Audio is part of the prompt, not a post step
Gemini Omni Flash renders dialogue, music, and sound effects natively, in the same generation as the picture. There is no separate audio pass and no lip-sync fix-up afterwards — you direct the soundtrack the same way you direct the camera: write it into the prompt. A line of spoken dialogue, a mood of music, a door slam, all from one text box.
Any duration from 3 to 10 seconds
Pick any whole-second length from 3 to 10. Pricing depends on the duration you choose, and longer clips can cost less in total: a 10-second 720p clip is $0.50 while a 5-second one is $0.63, because even durations (4, 6, 8, 10 seconds) hit the model's floor rate of $0.05 per second at 720p. When in doubt, generate the longer clip and trim.
Native 720p with 1080p and 4K upscaling
The model generates natively at 720p; 1080p and 4K options upscale that render for delivery. On 10-second clips, 1080p is $0.63 and 4K is $1.05 total. Both landscape 16:9 and vertical 9:16 are supported, so the same prompt can serve a widescreen cut and a story-format cut.
One node in a larger pipeline
On a Scenetra board, Gemini Omni Flash is a node you wire into a workflow: draft prompts with the copilot, generate a clip, then feed it onward — into the family's own Video Edit mode for a text-directed revision, or alongside other video models for comparison. Every run is billed at provider cost with 0% markup, pay-per-generation, no subscription gate.
Parameters
| Parameter | Type | Description | Default |
|---|---|---|---|
| Prompt* | text | Describe the scene, action, camera, style — and the audio: dialogue, music, and sound effects are rendered natively and directed purely through this prompt. | — |
| Duration | select | Duration of the video in seconds (3-10).3 · 4 · 5 · 6 · 7 · 8 · 9 · 10 | 8 |
| Resolution | select | 720p is native; 1080p and 4K upscale the 720p render.720p · 1080p · 4k | 720p |
| Aspect Ratio | select | Aspect ratio of the generated video.16:9 · 9:16 | 16:9 |
Pricing
From $0.05 per second
Pay per generation — only pay for what you use.
720p, 5s clip
$0.125/s
per second
720p, 10s clip
$0.05/s
per second
1080p, 10s clip
$0.063/s
per second
4K, 10s clip
$0.105/s
per second
What a video costs
| Duration | 720p | 1080p | 4K |
|---|---|---|---|
| 5 seconds | $0.63 | — | — |
| 10 seconds | $0.50 | $0.63 | $1.05 |
Use cases
Social content
- Sound-on hooks with dialogue
- Vertical 9:16 story clips
- Trend reactions with music
- Ambient mood loops
Marketing & ads
- Short spots with voiceover
- Product teasers with SFX
- Concept variants at 720p
- Finals upscaled to 4K
Film & previz
- Dialogue scene sketches
- Establishing shots with ambience
- Sound design exploration
- Beat-by-beat storyboarding
Education & explainers
- Narrated micro-lessons
- Concept visualizations with music
- Language practice clips
- Illustrated how-to beats
Frequently asked questions
What is reference-to-video in Gemini Omni Flash?+
It's a generation mode that takes up to 10 reference images — subjects, products, styles, compositions — and composes a new 3-10 second video that keeps them consistent. Unlike image-to-video, the references aren't the first frame; they're identity anchors the model draws from while building a new scene, with native audio directed through the prompt.
How is reference-to-video different from image-to-video?+
Image-to-video starts the clip from your exact uploaded frame and animates forward. Reference-to-video composes a brand-new scene while keeping the referenced subjects and style consistent — better when you want the same character or product in a different setting, angle, or action than your source photos show.
How many reference images can I use?+
Up to 10 per generation. Use them for whatever needs to stay consistent: a character from several angles, a product, a wardrobe, a location, or a visual style. The prompt tells the model how each reference should be used in the final scene.
Does it generate audio too?+
Yes — like every Gemini Omni Flash mode, dialogue, music, and sound effects are rendered natively in the same pass as the picture and directed through the prompt. Your referenced character can deliver a scripted line in the generated scene.
Related models
MiniMax H3
MiniMax H3 is the newest generation of the Hailuo video lineage: 5-15 second clips with native stereo audio — dialogue, music, and sound effects rendered in the same pass as the picture. Generate at 480P or 768P natively, or step up to 2K and 4K upscaled output, in six aspect ratios from cinematic 21:9 to vertical 9:16.
View model →Wan 3.0
Alibaba's Wan 3.0 generates up to 30 seconds of video in a single take, with native synchronized audio — dialogue, sound effects, and music — produced in the same pass as the picture. Pick any duration from 2 to 30 seconds, output at 480p, 720p, or 1080p, and choose from five aspect ratios or let the model frame the shot automatically.
View model →Kling 3.0 Standard
Kuaishou's standard-tier video generation model offering text-to-video and image-to-video with support for start/end frames, element references for character consistency, and native audio generation. A cost-effective alternative to Kling 3.0 Pro.
View model →Seedance 2.5
ByteDance's Seedance 2.5 generates cinematic videos up to 30 seconds from text prompts with synchronized audio — voice, sound effects, and background music — built in by default. The newest Seedance generation, with mp4 or high-fidelity mov output for multi-round editing pipelines.
View model →Start creating with Gemini Omni Flash Reference to Video
Use Gemini Omni Flash Reference to Video alongside 50+ other AI models in Scenetra's visual workflow editor. No setup required.
Get Started