Kling API: How Per-Second Billing Works, Which Tier to Pick, and How to Call It (2026)
A working guide to the Kuaishou Kling video API: how per-second billing turns a shot list into a budget, when the pro, audio and motion-control tiers are worth paying for, and how to call it without a Kling account.
1. What per-second billing changes
Most video tools sell you either a subscription with credits or a flat price per clip. Both hide the variable that drives cost — how long the clip is. Kling bills each second of requested video at the rate of the tier you pick, so the price of a job is simply rate × seconds, and you know it before you press submit. That makes budgeting a spreadsheet exercise rather than a guess: list your shots, give each a tier and a duration, and the column total is your render bill for one take of everything. The catch is the same as with any generative model — you will not keep every take. So multiply by your expected number of takes per keeper, and measure that number on a small batch first rather than assuming it.
2. Budget a shot list in three lines
Write the formula down once and reuse it for every project. Cost per usable second = tier rate × takes per keeper. Cost per shot = cost per usable second × final shot length. Project budget = sum over shots, plus a margin for re-shoots. The second line is where people overspend: they render every draft at final length. A ten-second shot drafted five times at full length costs five times the final render; drafted at the shortest useful length until the motion is right, then rendered once at full length, it costs a fraction. Work out takes per keeper by rendering 10–20 drafts of one representative shot and counting the ones you would actually use.
3. Standard vs pro: draft cheap, finish sharp
Standard and pro are the same model family at different quality settings. Standard is the right default while you are still discovering the shot — framing, timing, whether the motion reads at all. None of that needs pro detail. Pro earns its higher per-second rate on the take you will ship, where texture, faces and steadier motion are visible. A pattern that works: lock the prompt and the keyframe on standard, then re-render the winning prompt on pro at final length. If a prompt never works on standard, pro will rarely rescue it; fix the prompt or the image instead of paying more per second to discover the same problem in higher detail.
4. Audio tiers and motion control: when the surcharge pays
The audio tiers generate sound together with the picture, which matters when the sound has to be tied to what happens on screen — a door slam, footsteps, rain hitting a window. If you will lay a music bed or a voiceover over the edit anyway, generating audio is wasted spend; render silent and add sound in post. Motion control is for when free-form motion is not enough: a character has to perform a particular movement taken from a reference. It costs more per second than plain standard or pro, so use it on the few shots that need choreographed movement, not as a default. A useful test before paying for either: would a viewer notice if this shot were silent, or if the movement were approximately right? If not, stay on the cheaper tier.
5. Calling Kling through cocodot
POST https://cocodot.co/api/ai/video/generations with the Kling model name returned by the video model list (GET https://cocodot.co/api/ai/video/models), a mode of std or pro, a duration in whole seconds within the range we list, and your prompt (add a reference image for image-to-video). You get a job ID back straight away. Poll GET https://cocodot.co/api/ai/video/generations/{id} until the status is SUCCEEDED, then read the video URL and download it — generated links expire, so save the file to your own storage promptly. An accepted submission is not a delivered video: only the final polled state counts. The price is locked at submission and a failed job is not billed. The same key and balance call Hailuo and Seedance, so one poller covers all three vendors.
6. Consistency: let the image do the work
The most common complaint about AI video is that the character or product looks different in every shot. Text-to-video re-imagines the subject from words each time, so drift is expected. Image-to-video anchors the first frame, which is how you keep a face, a product or a brand palette stable across a sequence. Generate or design one reference image per subject, reuse it for every shot that subject appears in, and keep the prompt about motion and camera rather than appearance. Repeating appearance details in the prompt can fight the image and reintroduce the drift you were trying to avoid.
7. Going direct to Kling instead
Kuaishou runs its own Kling developer platform, and if Kling is the only video model you need and you can pay it directly, going direct is a reasonable choice — you get new model versions and parameters first. Check the official Kling developer platform for its current pricing, how credits are sold and what signup requires; those details change and we do not restate them here. A relay makes sense when you want several video vendors behind one key and one balance, when your local card is declined by the vendor's checkout, or when you would rather not prepay separate credit packs with each vendor. If none of those apply to you, go direct.
8. Mistakes that quietly double the bill
Four habits account for most overspend. Rendering drafts at final length — draft short, finish long. Defaulting to pro or audio tiers for exploration — upgrade only the take you keep. Retrying a failed or rejected prompt unchanged — it usually fails again for the same reason, so change the prompt or the image first. And batch scripts with no spending cap — a loop bug can submit hundreds of jobs overnight, so put a per-day limit in your own code and log every job ID with its tier and duration. For a side-by-side of Kling, Hailuo and Seedance on billing model and strengths, see /hub/ai-video-api-price-compare.
Kling tiers: what each one is for
| Tier | Use it for | Skip it when |
|---|---|---|
| Standard | Exploring ideas, storyboards, high-volume drafts | The shot is going into the final cut and detail matters |
| Pro | Final takes — finer detail, steadier motion | You are still deciding what the shot should be |
| Standard / pro with audio | Ambient sound or effects that must match the picture | You will score or sound-design the edit anyway |
| Motion control | A character must follow a specific reference movement | Free motion described in the prompt is good enough |
| Image-to-video (any tier) | Keeping a character, product or brand look consistent across shots | You have no keyframe and the look is not yet fixed |