Video Generation

Generate video as an async job. Model-specific constraints for LTX 2.3 and BytePlus ModelArk are validated before submission.

POST /v1/video/generations returns a job_id immediately. Poll GET /v1/jobs/{job_id} until it finishes. Every request needs a plm_ API key.

POST/v1/video/generations

Submit a job#

Parameter	Type	Required	Description
model	string	Optional	Video model ID. Defaults to `veo-3`.
prompt	string	Required	Video description. Up to 4,000 characters.
duration_seconds	integer	Optional	Requested duration. 2–20 seconds, default 8. Kling allows only 5 or 10, so specify it explicitly; LTX and ModelArk models apply narrower model-specific ranges.
resolution	480p \| 720p \| 1080p \| 1440p \| 4k	Optional	Output resolution. Defaults to `720p`.
generate_audio	boolean	Optional	Whether to generate audio. It defaults to false to avoid an unexpected audio-variant charge.
image	string \| null	Optional	Compatibility alias for the first image URL. It is combined with `images` while preserving order.
images	string[]	Optional	Input-image URLs. Combined with the image alias, there may be at most 9; only HTTP(S) is accepted.
videos	string[]	Optional	Input-video reference URLs. At most 3; only HTTP(S) is accepted.
audios	string[]	Optional	Input-audio reference URLs. At most 3; only HTTP(S) is accepted.
charge_amount_krw	integer \| null	Optional	Fixed KRW settlement amount. Normal API keys cannot use it; only first-party delegated keys may.

request

curl https://apirouter.pleum.ai/v1/video/generations \
  -H "Authorization: Bearer $PLEUM_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: seedance-shot-001" \
  -d '{
    "model": "dreamina-seedance-2-0-260128",
    "prompt": "A drone shot flying over a misty mountain valley at sunrise",
    "duration_seconds": 8,
    "resolution": "720p",
    "generate_audio": false,
    "images": ["https://example.com/first-frame.png"],
    "videos": ["https://example.com/motion-reference.mp4"]
  }'

submit + poll (Python)

import time, requests

BASE = "https://apirouter.pleum.ai/v1"
HEADERS = {"Authorization": "Bearer plm_...", "Content-Type": "application/json"}

# 1. Submit the job — returns job_id immediately.
resp = requests.post(f"{BASE}/video/generations", headers=HEADERS, json={
    "model": "dreamina-seedance-2-0-260128",
    "prompt": "A drone shot flying over a misty mountain valley at sunrise",
    "duration_seconds": 8,
    "resolution": "720p",
})
job_id = resp.json()["job_id"]

# 2. Poll until the job finishes.
while True:
    job = requests.get(f"{BASE}/jobs/{job_id}", headers=HEADERS).json()
    if job["status"] in ("succeeded", "failed"):
        break
    time.sleep(5)

print(job["result_url"] if job["status"] == "succeeded" else job["error"])

submit + poll (TypeScript)

const BASE = "https://apirouter.pleum.ai/v1";
const HEADERS = { Authorization: "Bearer plm_...", "Content-Type": "application/json" };

// 1. Submit the job — returns job_id immediately.
const submit = await fetch(`${BASE}/video/generations`, {
  method: "POST",
  headers: HEADERS,
  body: JSON.stringify({
    model: "dreamina-seedance-2-0-260128",
    prompt: "A drone shot flying over a misty mountain valley at sunrise",
    duration_seconds: 8,
    resolution: "720p",
  }),
});
const { job_id } = await submit.json();

// 2. Poll until the job finishes.
let job: any;
do {
  await new Promise((r) => setTimeout(r, 5000));
  job = await (await fetch(`${BASE}/jobs/${job_id}`, { headers: HEADERS })).json();
} while (job.status !== "succeeded" && job.status !== "failed");

console.log(job.status === "succeeded" ? job.result_url : job.error);

202 Accepted

{
  "job_id": "f47ac10b-58cc-4372-a567-0e02b2c3d479",
  "status": "processing",
  "model": "dreamina-seedance-2-0-260128",
  "result_url": null,
  "cost": null,
  "error": null,
  "assets": []
}

Provider-specific video input limits#

The selected LTX or ModelArk model's duration, resolution, and reference combination is validated before a hold is created. Unsupported combinations return 400 with no charge.

Parameter	Type	Required	Description
Kling	duration	Optional	Only 5 or 10 seconds are supported. Specify duration_seconds explicitly as 5 or 10, otherwise a 400 may be returned. Models: kling-v2-master, kling-v2-1, kling-v1-6, kling-v1-5, kling-v1-0.
LTX 2.3 Fast / Pro	duration + resolution	Optional	Text-to-video only. Fast supports 6/8/10/12/14/16/18/20 seconds at 1080p and 6/8/10 seconds at 1440p/4k. Pro supports 6/8/10 seconds at every resolution. 480p/720p requests normalize to LTX's 1080p output.
Seedance 2.0	duration + references	Optional	4–15 seconds. image≤9, video≤3, audio≤3. An audio reference needs at least one image or video reference. One/two images are first/last frames; with video, audio, or more than two images, every image is a reference_image.
Seedance 2.0 Fast / Mini	resolution	Optional	Only 480p or 720p. Mini does not support video or audio references.
Seedance 1.5 Pro	duration + references	Optional	4–12 seconds and no 4k. It accepts at most two image references and no video/audio references.
Seedance 1.0 Pro	duration + references	Optional	2–12 seconds and no 4k. At most two image references (one for Fast), no video/audio references, and no generate_audio support.

Poll a job#

GET/v1/jobs/{job_id}

POST initially returns status: processing. GET /v1/jobs/{job_id} returns the same job shape; a completed job can include video results in assets.

status is one of processing, succeeded, failed, or canceled. Success populates result_url/cost; failure or cancellation returns error. An expired upstream task is normalized to failed. Veo results are mirrored to the router's own storage, so result_url is a working presigned URL.

succeeded

{
  "job_id": "f47ac10b-58cc-4372-a567-0e02b2c3d479",
  "status": "succeeded",
  "model": "dreamina-seedance-2-0-260128",
  "result_url": "https://cdn.pleum.ai/video/f47ac10b-58cc-4372-a567-0e02b2c3d479.mp4",
  "cost": {"krw": 4200},
  "error": null,
  "assets": [{"type": "video", "url": "https://cdn.pleum.ai/video/f47ac10b-58cc-4372-a567-0e02b2c3d479.mp4"}]
}

Settlement and BYOK#

LTX video is held and settled from the requested duration at the selected resolution's per-second price. ModelArk video is finally settled from the completed task's reported completion-token/output usage. Idempotency-Key is optional (max 128 chars) and replays the same body for 24 hours.

Async generation such as video and 3D does not support BYOK. A BYOK-key request returns 400. A missing or not-owned job_id also returns 404.