Google Gemini Omni Flash · generate or restyle in seconds 
Gemini Omni Flash Prompt-first short clips and restyles
Text, a still, or a source video — 3–10 seconds, 16:9 or 9:16, 30 credits per second.
Run Gemini Omni Flash when you need a short clip fast or want to restyle footage you already have. Start from a prompt, animate a still, or send a source video with a natural-language edit. Set length first so credit math is honest, then generate on SupaImagine.
- Text-to-video drafts from a single prompt
- Image-to-video when a still already owns the look
- Source-video restyle by natural-language edit
- 3–10s duration · 16:9 or 9:16
- 30 credits per second on every path
Gemini Omni Flash: generate or restyle short video by prompt
Flexible 3–10s runs with a real source-video edit path — not a fixed eight-second only lane.
Text-to-video and image-to-video expose integer durations from 3 to 10 seconds (default often 8s) at 30 credits per second. Aspect is 16:9 or 9:16. The edit path ships your source video URL plus the prompt; billing follows the clip you upload. Paid plans cover commercial use under the Terms.
| Gemini Omni Flash | Omni Flash | Google Gemini Omni Flash | Gemini Omni Flash text to video | Gemini Omni Flash image to video | Gemini Omni Flash edit |
Three starts · one per-second rate
Three ways to start — one simple 30-credit-per-second rate
Pick the material you already have — blank prompt, still, or source clip — then pay only for the seconds you queue.
Write a short scene when nothing exists yet
Subject, action, camera, light, and optional sound cues go in one prompt. Text-to-video is for a continuous three-to-ten-second beat — hooks, inserts, product moments — not a multi-scene cut list. Negative instructions live in the same prompt (for example no on-screen text). You set 16:9 or 9:16 and the second count before the run, so exploration can start short and only lengthen when the take is worth keeping.
Animate a still that already locked the look
Upload a product shot, UI frame, illustration, or generated key art when identity is already decided. Describe the motion and camera you want; the still carries composition and materials while the prompt drives movement. The same 3–10s range and 16:9 / 9:16 choices apply, so a three-second motion check costs less than committing an eight- or ten-second pass on an unproven board.
Restyle footage you already shot or generated
Upload a source clip and write the change: mood, lighting, background, wardrobe, surface, or overall style. Keep the instruction focused — one primary change plus an explicit “keep everything else the same” when only one region should move. The edit path sends video URL and prompt upstream; length for billing follows the detected source duration rather than a free-form duration slider you invent. Use this when the camera move is already right and only the look needs a new pass.
Pay 30 credits per second — pick the length first
Text and image runs bill the same way: 30 credits times the seconds you select. A three-second probe is 90 credits, the eight-second default is 240, and a ten-second keep is 300. That linear rate is the point — you are not guessing a resolution tier for a fixed eight-second Google take. Signup balance is ten credits and cannot fund the ninety-credit floor, so treat the free grant as orientation, not a free multi-second clip.
From material to seconds
Prompt, still, or clip — then set seconds and frame
Four decisions before the queue: starting material, instruction, duration or billing length, and aspect.
1
Choose text, still, or uploaded clip
Blank brief for a new scene, a still when the board is already approved, or a source video when the job is restyle rather than first generation.
2
Write one clear instruction
For generation, spell subject, motion, camera, and optional audio cues. For edits, name the single change and what must stay fixed.
3
Lock seconds and 16:9 or 9:16
On text and image paths, pick 3–10 seconds and landscape or vertical. On edits, billing follows the uploaded clip length detected for the run.
4
Queue, review, then restyle or switch
Run the pass, check the returned file (including whether sound is present), then restyle again, lengthen a winner, or move the brief to Veo or Seedance when the job outgrows this model.
Same workspace, different first model
Omni Flash vs Veo and Seedance for this pass
Choose by whether you need prompt restyle, a fixed Google quality beat, or longer multimodal control.
| Veo 3.1 ~8s · Google quality family | Seedance 2 4–15s · multimodal refs | ||
|---|---|---|---|
| Best when you need | Generate or restyle short clips by prompt | A polished fixed-length Google-style beat | Longer timing plus multimodal references |
| Inputs on SupaImagine | Text; still; source video + edit prompt | Text; first frame; first+last frame | Text; frames; video / reference media |
| Duration exposed here | 3–10s on t2v/i2v; edit bills by source | ~8s fixed-class take | 4–15s |
| Credit posture | 30 credits per second | Resolution-tiered per clip | Duration- and tier-dependent |
| Skip this model when | You need lip-sync dialogue or long multimodal stacks | You need source-video restyle by prompt | You only need a three-second prompt restyle |
Jobs that fit this short-edit loop
When the next instruction is the edit
Situations where generate-or-restyle on a per-second rate beats a fixed-length quality pass or a heavy reference stack.
Vertical hook probes
Spin three-second 9:16 tests, then stretch only the winner toward eight or ten seconds before you commit a heavier model.
Packshot motion from a still board
Animate an approved product or packaging still when composition is locked and you only need a short motion pass.
Look restyles on existing footage
Upload a clip that already has the right move and ask for lighting, palette, background, or style without reshooting.
Before-and-after transformation beats
Show one focused visual change so viewers understand the edit, not a brand-new unrelated scene.
Campaign insert scouts
Prototype a single continuous insert on the cheap end of the per-second rate, then promote winners to a higher-quality or longer model.
Same brief, three model families
Run Omni Flash for restyle flexibility, Veo for a fixed Google-style beat, and Seedance when length and multimodal refs matter more.
Explore other video models
One plan, every video model — pick a better fit for motion, length, or style.
Veo 3.1
Text or first/last frames — not reference stacks, not flexible duration. Sound rides with the picture by default; lock 720p, 1080p, or 4K before credits move.
Seedance 2
The only Seedance tier with 1080p and 4K — always-on synced audio, 4–15s from text, frames, or refs.
HappyHorse 1.1
Joint audio and mouth motion in 7 languages — text, a first frame, or up to 9 reference images.
Veo 3.1 Fast
Text, first/last frames, or up to three stills — native audio on most takes, 720p to 4K from 30 credits.
Gemini Omni Flash questions before you generate
How long can a Gemini Omni Flash take be, and what does each second cost?
Text and image paths expose 3–10 second lengths at 30 credits per second — 90 credits for 3s, 240 for 8s, 300 for 10s. Pick duration before you run so the estimate matches the file you want.
What inputs does Omni Flash accept?
A written prompt for text-to-video, a still for image-to-video, or a source video plus prompt for restyle/edit. That edit path is the reason many teams open Omni Flash instead of a fixed-length only model.
What actually ships on a source-video edit?
The edit endpoint forwards the source video URL and your prompt. Duration is used for credit math based on the uploaded clip rather than inventing a free-form rewrite length. Plan rights to any footage you upload.
Should I expect audio on every Omni Flash file?
Treat audio as common when the model returns scored motion, but do not assume every restyle keeps a perfect music bed. Review the file before client handoff.
When is a 3-second run the right call?
Use the shortest length for rapid look tests and prompt hunting. Stretch toward 8–10 seconds when the beat needs a fuller arc and you are ready to spend the per-second cost.
How is Omni Flash different from a fixed ~8s Google Veo clip?
Omni Flash lets you choose 3–10 seconds and restyle an existing clip by prompt. Veo quality/Fast lanes on this site are about eight seconds per run with different input and pricing rules.
Which aspect ratios are available?
16:9 and 9:16 on the paths exposed here. Set the ratio with the brief so vertical and horizontal stories stay intentional.
Are paid-plan Omni Flash clips cleared for commercial work?
Paid-plan runs follow SupaImagine Terms for commercial use, including prompt-restyled source clips — you still need rights to any third-party footage, talent, or audio you upload.
Before and after the Omni pass
Pair Omni Flash with still boards and finish tools
Build the still first, restyle the motion here, then upscale or switch models without leaving the product.
Next short clip: generate it or restyle it with Gemini Omni Flash
Text, still, or source video in — 3–10s at 30 credits per second — check the file, then iterate.