How to Use Gemini Omni Image-to-Video: 4K Clips from One Photo
A full guide to Gemini Omni Image-to-Video on VdoBloom: real durations, 720p/1080p/4K credit costs, why 720p is never worth picking, prompting structure, and when Kling 3.0 Omni or Wan 3.0 wins instead.
Gemini Omni Image-to-Video on VdoBloom animates a still photo into a 4, 6, 8 or 10 second clip at 720p, 1080p or 4K, costing 36 to 72 credits below 4K and 84 to 120 credits at 4K — so a 10-second 4K render is $6.00 and a 4-second 720p test is $1.80. It is one of the few image-to-video models on the platform with a true 4K output tier, and at 120 credits for a 10-second 4K clip it is less than half the price of Kling 3.0 Omni at the same spec.
What the model is best at
This is Google’s Gemini Omni video engine running through VdoBloom’s credit system, in its image-to-video configuration. You supply a start frame and a prompt describing the motion; the model extends outward from your composition rather than inventing a scene. That makes it a finishing tool rather than an exploration tool. It suits work where the still is already approved: a product render, a shot photograph, an illustration, an archival frame.
The reason to choose it over cheaper animators is the resolution ceiling. Most image-to-video models stop at 1080p, which is fine for a phone feed and thin on a large screen or in a paid placement. Gemini Omni renders at 4K, and the price step up to 4K is modest — 48 extra credits at ten seconds, not a doubling.
Users rate it 4.7, the joint highest rating in the family alongside its text-to-video sibling.
Real specs
| Spec | Gemini Omni Image-to-Video |
|---|---|
| Durations | 4, 6, 8 or 10 seconds |
| Resolutions | 720p, 1080p, 4K |
| Aspect ratios | 16:9 and 9:16 only |
| Input | A start image plus a text prompt |
| Reference image ceiling | Up to 7 images (20 MB each) at the provider level |
| Seed control | Supported, for repeatable renders |
| Content filter tier | STRICT |
| User rating | 4.7 |
There is no 1:1, 4:3 or 21:9 option. The two ratios cover horizontal web and vertical social, and that is deliberate — if a brief needs square framing, the model is the wrong pick and no crop workaround gets you a native square render.
Exact credit cost per tier
One credit is $0.05 at the Lite plan rate of $15 for 300 credits, so these dollar figures are exact.
| Duration | 720p / 1080p | Dollar cost | 4K | Dollar cost |
|---|---|---|---|---|
| 4 seconds | 36 credits | $1.80 | 84 credits | $4.20 |
| 6 seconds | 48 credits | $2.40 | 96 credits | $4.80 |
| 8 seconds | 60 credits | $3.00 | 108 credits | $5.40 |
| 10 seconds | 72 credits | $3.60 | 120 credits | $6.00 |
Read that table carefully, because it contains a genuinely useful quirk: 720p and 1080p cost exactly the same. There is no reason to render this model at 720p. If you are not going to 4K, always choose 1080p — identical credits, better output. The only situation where 720p makes sense is if your pipeline specifically requires it.
The second thing worth knowing: the sibling Omni Flash 1.1 variant adds a 360p tier, and it is priced identically to 720p and 1080p at every duration. A cheap-looking resolution option that is not cheaper is not a saving, so ignore it.
At 12 credits per second of 1080p output, a 10-second clip is 72 credits. The 300-credit Lite plan is therefore four 10-second 1080p clips a month, or two 10-second 4K masters with 60 credits left over. Full plan maths is on the pricing page, and annual billing brings Lite down to $10 a month.
How it compares to the other 4K-capable animators
| Model | 10s at 1080p | 10s at 4K | Max duration |
|---|---|---|---|
| Gemini Omni Image-to-Video | 72 credits ($3.60) | 120 credits ($6.00) | 10 seconds |
| Kling 3.0 Omni Image-to-Video | 108 credits ($5.40) | 268 credits ($13.40) | 15 seconds |
| Wan 3.0 Video | 128 credits ($6.40) | No 4K tier | 30 seconds |
Gemini Omni is the cheapest route to a 4K animated still on VdoBloom by a wide margin. What it does not have is length: ten seconds is the ceiling, where Kling 3.0 Omni reaches fifteen and Wan 3.0 reaches thirty. It also carries the STRICT content label, VdoBloom’s tightest tier, while Wan sits in the flexible HOT tier that handles swimwear, dance and fitness material mainstream tools refuse.
How to prompt it well
Because the first frame is fixed, the prompt should describe change over time, not the image itself. Describing what is already in the photo wastes the prompt and sometimes causes the model to re-render features you wanted preserved.
- Give one camera instruction. A slow push-in, a lateral dolly, a gentle orbit. Two competing camera moves in ten seconds reads as drift.
- Give one subject action. Steam rising, fabric settling, a head turning to camera, a hand lifting the product.
- State what holds still. Naming the background, lighting and identity as unchanged materially reduces morphing on faces and logos.
- Match the motion to the duration. A four-second clip supports one beat. A ten-second clip supports a beginning, a change and a settle — write all three.
- Reuse the seed. Once a render is close, keeping the seed and changing one prompt clause isolates the variable instead of rerolling the whole clip.
The exact steps
- Open the Image to Video tab and pick Gemini Omni Image-to-Video from the model picker.
- Upload your start frame. Sharp, well-lit, with the subject fully in frame — the model amplifies whatever softness is already there.
- Choose 16:9 or 9:16 to match where the clip will run, and select 1080p unless you specifically need 4K.
- Set duration. Draft at 4 seconds for 36 credits before committing 120 to a 4K ten-second master.
- Write the motion prompt using the camera-plus-action structure above.
- Check the credit cost shown before you generate — VdoBloom prices every duration and resolution combination up front, so there are no surprises.
- Generate, review, then re-run the winning prompt at your delivery resolution.
If the photo shows a real person, you must have that person’s consent before animating them. The model is also STRICT, so uploads involving sensitive real-person material may be refused at the provider level before any render starts.
When to pick a different model
Pick Kling 3.0 Omni when you need eleven to fifteen seconds in one shot and the budget carries the 4K premium. Pick Wan 3.0 Video when you need long-form output, native audio, or a more permissive content tier. Pick a faster, cheaper model entirely when you are exploring concepts rather than finishing them — at 36 credits a draft, Gemini Omni is not the cheapest place to experiment. Full specs sit on the Gemini Omni Image-to-Video model page.
Frequently asked questions
How much does a 10-second 4K Gemini Omni clip cost?
120 credits, which is $6.00 at the Lite rate of $15 for 300 credits. The same length at 1080p is 72 credits, or $3.60.
Is 720p cheaper than 1080p on this model?
No. Both tiers cost the same credits at every duration — 36, 48, 60 and 72 for 4, 6, 8 and 10 seconds. Always render 1080p rather than 720p.
What is the longest clip it can produce?
Ten seconds. There are four fixed options: 4, 6, 8 and 10 seconds. For longer single renders, Kling 3.0 Omni reaches 15 seconds and Wan 3.0 Video reaches 30.
Can I make a square video with it?
No. The model outputs 16:9 and 9:16 only. A different model from the catalogue is the right route for 1:1 or 4:3 work.
Will my videos have a watermark?
Paid plans download watermark-free with commercial rights. New accounts get 10 free credits without a card, which is not enough for a single Gemini Omni render — the cheapest tier is 36 credits — so this is a model you reach after topping up, with one-time credit packs starting at $2.49 for 75 credits that never expire.
Ready to try it?
Create your first AI video in minutes — no credit card required.
Start Creating Free →