Kling 3.0 shot timing prompts: make seconds add up on Sume
Kling's guide writes multi-shot prompts as 'Shot 1, 4s'. On Sume kling-3 has no shot list field, so the timing lives in the prompt and in duration.

How does Kling say to write a timed multi-shot prompt?
Write each shot on its own line with its length, and make the shot lengths add up to the clip length you chose. Kling's own narrative guide gives the pattern as "Shot 1, 4s: [description]. Shot 2, 6s: [description]." and says the total must equal the output length you pick.
The same page says Kling's Custom Multi Shot mode supports up to 6 distinct shots and a 3 to 15 second output, and it recommends cinematic wording such as "slow dolly push" or "shaky handheld" plus audio notes (dialogue, effects, music) in each shot. Allocating longer shots to reflective beats and shorter ones to urgency is the guide's pacing advice.
That is a description of Kling's own product surface. What matters for a Sume buyer is which parts of it reach the kling-3 model id.
What does Sume's kling-3 accept?
Sume lists kling-3 in its video catalog with text-to-video, image-to-video with first and last frame, optional audio, 720p or 1080p, aspect ratios 16:9, 9:16 and 1:1, and whole-second durations from 4 to 15. The request has a prompt, a duration integer and the other fields in the video generation docs; there is no shot list, no per-shot duration field and no Kling-specific passthrough, because provider.options is rejected for every model in v1.
So the shot structure can only travel as text inside prompt. Whether Kling follows "Shot 2, 6s" literally for a given request is not something the Sume docs promise, and this post does not either. Treat the timed prompt as a strong hint and check the result.
| Item | Kling narrative guide | Sume kling-3 |
|---|---|---|
| Output length | 3 to 15 seconds | 4 to 15 seconds, whole seconds |
| Shot count | Up to 6 shots in Custom Multi Shot | No shot field; shots are prompt text |
| Shot lengths | Must add up to the chosen length | You keep the sum equal to duration |
| Audio notes | Dialogue, effects, music per shot | Prompt text plus the generate_audio flag |
What does a 10-second, two-shot request look like?
Pick the total first, then split it. A 10 second clip with a 4 second establishing shot and a 6 second payoff shot is a duration of 10 and a prompt whose two shot lengths sum to 10.
Send it to POST /v1/videos with a bare catalog id, no provider prefix. Replace the placeholder text with your own scene.
curl -X POST "https://api.sume.com/v1/videos" \
-H "Authorization: Bearer $SUME_API_KEY" \
-H "Content-Type: application/json" \
-H "Idempotency-Key: kling3-two-shot-001" \
-d '{
"model": "kling-3",
"prompt": "Shot 1, 4s: wide shot of a bakery counter at dawn, slow dolly push. Shot 2, 6s: close-up of a hand dusting flour on a loaf, soft steady light.",
"duration": 10,
"resolution": "720p",
"aspect_ratio": "16:9",
"generate_audio": false
}'Which lengths break the rule?
Three limits catch people who copy the Kling guide straight across.
First, a 3 second clip is inside Kling's stated range but outside the Sume catalog row, which starts at 4 seconds. Second, a total above 15 is outside both. Third, the Sume Videos panel offers Kling 3.0 durations of 5, 6, 8, 10 and 15 seconds only, so a 7 or 9 second plan is an API-only length; see the panel versus API comparison.
- Do the arithmetic before you submit: the shot seconds in the prompt should equal
duration. - Stay at 4 to 15 whole seconds for
kling-3; readsupported_durationsfromGET /v1/videos/modelsinstead of hard-coding. - Send an
Idempotency-Keyso a retried submit returns the original job instead of a second charge.
What does Sume not do here?
Sume does not enforce or validate the shot lengths in your prompt; it only validates duration against the model row. It also does not stitch separate shots for you inside one job. If you need hard cuts at exact seconds, the reliable pattern is one clip per shot and an edit on the timeline, covered in Kling multi-prompt shots versus one clip per shot.
Billing follows the model row: the workspace balance is reserved on submit at the provider list rate times 1.25, and usage.cost on the finished job is the Sume billable amount. A longer clip reserves more, so confirm the total before you ask for the full 15 seconds.
Sources
Related posts
More in Models
- Kling 3.0 subject binding limits and what Sume kling-3 takes
Kling's subject binding takes up to 4 images, a 3-8 s clip and a voice sample. Sume's kling-3 takes none of them, so lock identity with a first frame.
- Kling 3.0 in the Sume Videos panel: 5-15 s choices vs API 4-15 s
The panel offers Kling 3.0 at 5, 6, 8, 10 or 15 seconds and no end frame. The API accepts any whole second from 4 to 15 and a last frame.
- Kyutai Pocket TTS languages: what it speaks vs hosted Sume TTS
Pocket TTS is a 100M-parameter open model you run yourself. Its README lists seven languages; Sume's TTS is hosted, per character, with a voice id.
- Longest AI video clip in one request: 30, 15 or 10 seconds by model
Seedance 2.5 and Wan 3.0 reach 30 seconds on Sume; most other rows stop at 15 and Gemini Omni Flash at 10. Ceilings per model, and when to stitch instead.
Written by Sume