kling Video generation
Kling V3
Plan the shot. Direct the motion. Review the take.
Kling V3 is TokenHot’s route for Kling Video 3.0. Explore short product scenes, narrative sequences and dialogue concepts with a brief that names the subject, action, camera movement and sound. Start with a small, reviewable shot and judge motion, continuity and timing before combining accepted clips into a finished edit.
- Video output
- Shot direction
- Audio option in API examples
This page currently provides API access guidance without an embedded Kling playground. Video 3.0 and Video 3.0 Omni are distinct models with different controls.
USD /s
Pricing
| Group | Tier | Price |
|---|---|---|
| default | 720P | $0.09 |
| default | 有声，4K | $0.09 |
| default | 无声，1080P | $0.1197 |
| default | 有声，720P | $0.135 |
| default | 有声，1080P | $0.135 |
| default | 无声，4K | $0.45 |
Public list prices are shown in USD. Final charges may vary by account group and usage tier.
Overview
What is Kling V3?
Kuaishou introduced Kling Video 3.0 as part of its February 2026 model release. Its Video 3.0 materials describe native audio, multi-shot generation and clips up to 15 seconds. These are provider capabilities; the accepted settings of the TokenHot route must be checked separately.
Direct a video as a sequence of visible events. Specify who or what is in the frame, how it moves, where the camera goes and what should be audible. Keep the brief small enough that you can identify why a take succeeds or fails, then use an editor to assemble and finish the sequence.
Make the action explicit
Describe the subject’s movement and the camera movement separately. Use a simple starting shot to expose continuity or geometry problems.
Give each shot a purpose
Specify the opening, action and final frame. Decide what must remain consistent across a cut before adding more shots.
Write an audible brief
Name the spoken line, speaker and background sound. Keep dialogue concise enough to review timing and pronunciation clearly.
From brief to deliverable
Put Kling V3 to work
Brand & performance creative teams
Develop a short product reveal
Build a single product shot around one reveal: a camera push, a controlled turn or a change in light. Start with a precise description, or a reference through a route that documents image input. Add final titles and exact brand marks during editing.
- Tools you need
- The video API and a video editor. Reference-image submission requires the documented input fields for your route.
- Keep in mind
- Inspect the product throughout the clip, not just the first frame. Avoid treating generated geometry or logos as exact.
View a task brief
Create a short product reveal of a matte ceramic travel cup on a neutral studio surface. Begin with a close view of its texture, then make one slow camera pullback to reveal the whole cup. Keep its geometry, handle and color consistent throughout. Use soft side lighting and a quiet background. Add no text, logos or people. End on a stable composition with space for a separate end card. Review for shape drift, flicker and unintended camera motion.
Directors & content studios
Test a scene before committing to a full edit
Use a short sequence to explore whether a narrative beat reads clearly. Describe a small number of shots, each with a defined role. A rough scene can help a team discuss pacing and continuity before spending time on the final production.
- Tools you need
- The video API for generation and an editor for trimming, sequencing and review.
- Keep in mind
- Provider multi-shot capability does not imply that Omni-specific storyboard fields are accepted by this route.
View a task brief
Visualize a brief scene in which a cyclist notices a dropped notebook, stops and returns it to its owner. Use an establishing view, a closer view of the discovery and a final exchange. Keep clothing, bicycle color and notebook appearance consistent across cuts. Use natural daylight and restrained camera movement. Avoid dialogue and on-screen text. The scene should pass review only if the sequence is easy to follow and the handoff contains no duplicated objects or abrupt identity changes.
Social video & narrative creators
Prototype a dialogue moment with sound
Keep the first audio experiment to one speaker and one short line. Describe delivery, surrounding ambience and camera framing. Review the sound independently from the image before judging the combined take or expanding to multiple characters.
- Tools you need
- A route with the documented audio option and a video editor for sound review or replacement.
- Keep in mind
- Confirm audio and duration settings before submission. Pronunciation, timing and lip synchronization require review.
View a task brief
Create a short scene of a shopkeeper standing inside a quiet flower shop. In a calm, welcoming voice, the shopkeeper says exactly: “Let’s find something for your table.” Keep the camera at eye level with one gentle push forward. Use subtle room ambience and no music or additional voices. Leave a short pause before and after the line. Accept the take only if the words are correct, speech is clear and the visible performance is consistent with the audio.
Capabilities in context
Key features
| Feature | What you get | Why it matters |
|---|---|---|
| Model family | Kling Video 3.0, distinct from Video 3.0 Omni and Image 3.0. | Do not transfer Omni reference or editing controls to this route without documentation. |
| Provider duration | Kuaishou describes video generation up to 15 seconds. | This is a provider maximum. Select only durations accepted by your TokenHot API route. |
| Shot direction | Provider materials describe multi-shot storytelling. | Specify shot intent in the brief; confirm whether separate storyboard parameters are available. |
| Native audio | Kuaishou describes generated speech and sound; TokenHot examples include an audio option. | Treat audio as part of both the creative brief and the billing configuration. |
| API request | TokenHot’s public example uses duration, size, aspect_ratio and audio_generation. | Check the API docs for authentication and accepted values before implementing a production client. |
| Asynchronous generation | TokenHot documents a task_id response followed by a video-task query. | Store the task identifier and poll its status instead of submitting the same generation again. |
| First and last frames | TokenHot documents image inputs through file_infos with FirstFrame and optional LastFrame usage. | Use the image-to-video request format; an image URL placed in ordinary prompt text is not a reference attachment. |
Provider capabilities and TokenHot example fields are separate evidence. Catalog presence and sample requests do not guarantee a particular configuration succeeds. Kling Video 3.0 · Kuaishou’s model announcement · TokenHot video API · First and last frames · Task query
Audience & fit
Where Kling V3 fits
Creative teams testing a shot
Use an explicit review rubric for continuity, motion and composition. Change one part of the brief at a time so a new take teaches you something.
Studios planning a sequence
Develop individual narrative beats, then assemble approved clips in an editor. Keep reference descriptions consistent across requests.
Developers integrating video generation
Build around the documented task lifecycle. Track each request, expose failures clearly and avoid silently resubmitting a costly generation.
Cost context
Plan the cost of a useful result
Match the unit to the job
TokenHot’s public configuration uses a per-second billing unit for Kling V3. Estimate using the applicable rate and billable duration, then include the number of candidate takes you expect to generate.
Resolution and audio combine
The public pricing expression applies conditional factors for resolution and sound. A single displayed option may not represent the combined configuration. Confirm the combined rate before budgeting a production batch.
Compare finished, accepted footage
Keep duration, resolution and audio settings the same when comparing options. Include rejected takes and post-production time; provider subscription credits do not translate directly into TokenHot API dollars.
Use the current pricing table for available rates and account groups. Model IDs and prices come from TokenHot’s catalog; provider pricing and subscriptions are separate.
API
Code examples
curl --location 'https://api.tokenhot.ai/v1/video/generations' \
--header 'Content-Type: application/json' \
--data '{
"model": "kling-v3",
"prompt": "A white long-haired cat lounges on a wooden windowsill in the sun, blooming roses in the background; the cat blinks slowly and gently flicks its tail",
"duration": 5,
"size": "1080P",
"aspect_ratio": "16:9",
"audio_generation": true
}'FAQ
Frequently asked questions
What is Kling V3 best suited for?
Kling V3 is suited to the input, output, and capability types listed on this page. Test production workloads before choosing it for a critical system.
How is Kling V3 priced?
Pricing is shown in the pricing section above. Actual cost depends on usage volume, account group, and the request payload.
How can I call Kling V3?
Use the model ID shown above with one of the supported API protocols. Code examples are provided when usage data is available.
Is Kling V3 the same as Kling V3 Omni?
No. Kuaishou presents Video 3.0 and Video 3.0 Omni separately. This page covers the Kling V3 route. Verify Omni-specific reference, storyboard and editing controls against the separate model’s documentation.
Where is the Kling V3 playground?
This detail page currently offers pricing, code examples and API documentation without an embedded Kling trial. Use the documented integration path and check the settings available to your account.
Can a Kling V3 clip include audio?
Kuaishou describes native audio for Video 3.0, and TokenHot’s public example includes audio_generation. Confirm the accepted configuration and its rate in the API docs. Review spoken words, ambience and synchronization in the generated result.
Can every request generate a 15-second clip?
Fifteen seconds is the provider-described maximum, not a guarantee for every route or mode. Confirm allowed durations and their billing conditions before submission.
Does a 4K pricing label prove native 4K output?
No. A billing label describes a configured pricing condition. It does not establish the model’s native output resolution, processing method or availability. Confirm the documented output for the exact request configuration.
How should I budget a video campaign?
Choose duration, resolution and audio first, then check the combined rate for your account. Multiply by the expected number of takes and include finishing work. Budget around accepted footage rather than assuming every generation will be usable.
Get started
Build with Kling V3
API