Skip to content

Video

Kling 3.0 Turbo

Video

Available on Ofrai — chat, images, and video in one place with shared credits.

366 creditsModel details

Video studio

Preview

No clip yet

Your latest clip preview will show here after generation.

Open studio

Example prompts

Choose an example to load its prompt and recommended settings into the Kling 3.0 Turbo creator above.

How it works

Create with Kling 3.0 Turbo in three steps

  1. 01

    Choose your inputs

    Keep Kling 3.0 Turbo selected, then add a prompt and any supported reference assets.

  2. 02

    Set the output

    Choose the aspect ratio, resolution, and available Video options before generating.

  3. 03

    Generate and refine

    Review the result, adjust the prompt or settings, and generate another version from the same place.

Kling 3.0 Turbo runs on the new Turbo protocol and is built around a single idea: get a clean, short clip out quickly, either invented from text or anchored to one still image you supply. There is no slot bookkeeping here — no numbered reference lists, no placeholder syntax, no tail frame — and the model will not quietly invent any of them on your behalf.

Output resolution is an ordinary parameter with two stops rather than a mode you have to reason about, and duration runs second by second all the way to fifteen. What that simplicity costs is equally concrete: a single still is the only anchor you get, so anything that needs an identity held across several takes, a defined ending frame, or native sound is a different model's job.

What Turbo is actually tuned for

Read the interface description literally and it tells you the model class: an asynchronous endpoint that accepts text-to-video and image-to-video, where image input supports the first frame only. There is no third input type. That is the whole surface, and it is a coherent one — Turbo is tuned for the shot you have already decided on, where the remaining unknown is motion rather than identity.

A product clip that opens on a frame you art-directed, a title card sequence, a short beat of a wider scene that you will assemble yourself later: these are jobs where you already know the composition and want the model to spend its effort on how the frame moves.

The protocol detail that matters in practice is that for image-to-video your text is converted into the prompt item inside the Turbo contents structure, so the same text field serves both modes and is interpreted as a description of movement relative to the frame you uploaded, not as a replacement for it.

Where it visibly fails, and why

The documented gaps are all limits of scope rather than quality bugs, and knowing them in advance is most of the work. First, image-to-video anchors the first frame and nothing else: you cannot give it a start and an end, so a shot that must land on a specific closing image is out of reach, and a cut that should arrive somewhere else has to be re-authored.

Second, aspect ratio is documented as applying to text-to-video only, which means that when you upload a still, the geometry comes from the image rather than from the parameter — asking for a vertical frame while supplying a landscape still does not re-compose the shot, and the mismatch shows up as a cropped or letterboxed result rather than an error.

Third, with no reference lists there is nothing holding a face, a product, or a costume stable across two separate renders, so multi-shot continuity has to come from your edit. Fourth, there is no audio parameter on this entry, so any sound is added downstream.

Prompt and input structure this model responds to

One required text field, used for both modes, and the shape of a good prompt differs sharply between them. For text-to-video the prompt carries everything — subject, action, setting, camera, style — because nothing else is supplied. For image-to-video the frame is already fixed, so the highest-value sentences are the ones about time: what moves, in which direction, at what pace, and where the camera goes.

Repeating what is already visible in the uploaded still is not neutral; it competes with the image for the model's attention and is a common cause of a first-frame composition that drifts a few seconds in. Write the subject's action and the camera move as separate sentences, keep each to one idea, and put the style words last so they read as a modifier rather than as an instruction competing with the action.

Because there are no slots, there is nothing to number and no order to keep consistent, which is exactly why this model is quick to iterate on: a second attempt is a re-roll, not a rebuild.

Parameters and the trade-offs they force

Three parameters do all the work. Resolution takes 720p, the default, or 1080p — a plain quality choice with no mode field attached, so there is no hidden second decision riding along with it the way a mode switch carries billing. Duration accepts every whole second from 3 through 15, with 5 as the default, which makes this the most precisely timeable entry of the Kling group. Aspect ratio takes 16:9, 1:1, or 9:16, defaulting to 16:9, but only governs text-to-video.

The real trade-off is length against coherence: seconds are billed individually, so a fifteen-second request costs three times a five-second one, and past roughly ten seconds you are asking one continuous take to hold a subject, a background, and a camera move for longer than single-take generation reliably manages. The sensible pattern is to pay for length only when the take is genuinely continuous, and to cut scenes into five-to-eight second clips and assemble them otherwise.

When to pick a sibling instead

Choosing Turbo because it has the newest number in the family is the most common way to overpay for it, so it is worth naming the cases where a neighbour is simply the better instrument. If the brief contains reference material that must survive into the output, the Omni entry in the same line gives you numbered image, video, and element lists with placeholder references, at the cost of index discipline.

If the deliverable needs sound, an explicit start and end frame, or a shot fixed to two standard lengths, the v2.6 entry pairs standard 720p silent output with a professional tier that generates audio automatically. If the job needs a reference video edited rather than a still anchored, the Video O1 entry documents that path, along with a reasoning-enhanced quality pass.

And if the requirement is multi-camera consistency across many reference images, the Vidu Q3 family is built for that, with a base model that takes up to seven references.

Moving an existing job onto Turbo

Two migrations are common, and each trades something specific. Coming from v2.6, you gain per-second timing across a 3-to-15 range instead of two buckets, and you lose automatic audio plus the mode field that decided resolution and sound together — the job now either carries no sound or gains sound in another pass, and your 1080p decision is a separate parameter rather than a mode upgrade.

Coming from a reference-driven Omni workflow, you lose the input lists and the placeholder syntax entirely and keep only the first-frame anchor, so prompts that addressed slots have to be rewritten as prose about a still. Checklist before switching: confirm the shot genuinely needs no second reference, confirm the still you will upload already matches the ratio you want since the parameter will not correct it, and confirm nothing in the brief depends on a closing frame.

FAQ

Does aspect_ratio apply when I generate from an image?
No. The parameter is documented as applying to text-to-video only. When you supply a still, the frame geometry comes from that image rather than from the requested ratio, so supply an image whose own proportions match your target.
Can I give Turbo both a start and an end frame?
No. Image-to-video on this entry supports the first frame only. If the shot has to arrive on a specific closing image, that requirement has to move to a model with explicit first-and-last frame control.
Which durations are available?
Every whole second from 3 to 15 is selectable, with 5 as the default. Longer requests cost proportionally more and are more prone to drift late in the take, so cut scenes into shorter clips when continuity is not the point.
Is a prompt required, and how is it used for image input?
Yes, the prompt field is required. For text-to-video it describes the whole shot; for image-to-video it is converted into the prompt item of the Turbo contents structure, so it should describe motion relative to the frame you already supplied.

FAQ

Can I use Kling 3.0 Turbo directly on this page?
Yes. The creator above is the working Ofrai studio with Kling 3.0 Turbo selected by default.
Where are my finished results saved?
Signed-in generations are kept in your Ofrai library so you can download or revisit them.
Which prompts work best with Kling 3.0 Turbo?
Describe the subject, setting, visual style, lighting, and camera or motion clearly. The examples above provide practical starting points.
Can I edit the example prompts before generating?
Yes. Load any example, then change its wording, aspect ratio, references, or other supported settings in the creator.

Compare other Video models that can handle similar creative work to Kling 3.0 Turbo.

Commercial use depends on the selected model and model terms.