--- title: "Kling 4.0" description: "Kling 4.0 makes video from text, from one first frame, or from reference media. It renders at 720p or 1080p, in any whole length from 3 to 30 seconds. It is the Kling model for new video: it keeps a person or a product the same from shot to shot." url: https://docs.roasos.com/models/video/kling-4 --- # Kling 4.0 Family `video` · model `kling-4` · catalog `model-catalog@2026-10-06.v5` · pricing `customer-pricing@2026-11-05.v15` . ## Endpoint ```http POST https://api.roasos.com/v1/video/kling-4 ``` Send your API key as a bearer token. The key needs the scope `api:write`. Every paid request also needs an `Idempotency-Key` header. The envelope fields `quote_only`, `quote_id`, `webhook_url` and `wait_timeout_seconds` behave the same on each model endpoint. Refer to [Jobs](/concepts/jobs) and [Quotes and credits](/concepts/quotes-and-credits). ## Good at Kling 4.0 is the only Kling video model since 2026-10-29. It replaced Kling 3, Kling 3 Pro, Kling O3, Kling O3 Pro and both 4K tiers, and those endpoints are removed. Use it to keep one or more people, or a product, the same across a clip. Send each person or product as an element: one frontal image and up to 3 more views. The prompt names each element once as `@Element1`, `@Element2`, and so on. Send `avatar_id` to attach a saved person as an element. An element can also be a video of the person or object (`video_url`, 3 to 8 seconds, at least 700 px on each side) instead of images. Up to 3 elements can be video elements, and their seconds count in the 30-second clip total and in the price. Up to 3 elements can carry a `voice_id`: a Kling voice that you made with fal's create-voice endpoint. The element then speaks in that voice. To change an existing video, use [Kling 4.0 Edit](/models/video-edit/kling-4). The reference lane takes up to 10 images (`references`, named `@Image1..N`), 5 video clips (`video_urls`, named `@Video1..N`) and 7 elements. A first frame (`image_url`) on the reference lane is the literal opening frame, and it counts as one of the 10 images. Each reference clip must be 3 to 30 seconds long, and all clips together must be 30 seconds or less. A reference clip shows motion or style to follow. It does not edit that clip. Use `multi_prompt` for a board of shots in one generation: up to 8 shots, each with its own prompt (512 characters or less) and its own length. The shot lengths must add up to `duration_seconds`. Text-to-video takes 9:16, 16:9, 1:1 and 21:9. The reference lane also takes `auto`, which follows the references. Image-to-video takes its framing from the first frame, and `end_image_url` sets the last frame. The provider documents no audio switch. The endpoint does not accept `generate_audio`. ## Pricing The unit is one **clip**. Pricing key: `kling-4`. | Request | Credits | | ----------- | ------- | | `3s_1080p` | 9 | | `17s_1080p` | 48 | | `30s_1080p` | 85 | The credits come from the roasOS pricing table. No other field changes the price. To get the exact price of your request, send it with `quote_only: true` first. ## Modes * `text-to-video` * `image-to-video` * `reference-to-video` roasOS selects the mode from the media fields of your request. You do not send a mode. Provider endpoints: `kling/v4/image-to-video` · `kling/v4/reference-to-video` · `kling/v4/text-to-video`. ## Request fields | Field | Type | Required | Default | Rule | | ------------------------- | ------------ | -------- | ------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------- | | `aspect_ratio` | enum | no | `16:9` | `9:16`, `16:9`, `1:1`, `21:9`, `auto`; only in `text-to-video`, `reference-to-video` | | `avatar_id` | string | no | — | UUID; only in `reference-to-video` | | `avatar_outfit` | string | no | — | ≤ 500 characters | | `duration_seconds` | integer enum | no | `5` | `3`, `4`, `5`, `6`, `7`, `8`, `9`, `10`, `11`, `12`, `13`, `14`, `15`, `16`, `17`, `18`, `19`, `20`, `21`, `22`, `23`, `24`, `25`, `26`, `27`, `28`, `29`, `30` | | `elements` | object\[] | no | — | 1–7 items; each `{ frontal_image_url, reference_image_urls, video_url, voice_id }`; only in `reference-to-video` | | `end_image_url` | string | no | — | https URL; ≤ 4,096 characters; only in `image-to-video` | | `extend_of_generation_id` | string | no | — | UUID | | `image_url` | string | no | — | https URL; ≤ 4,096 characters; only in `image-to-video` | | `mode` | enum | no | — | `text-to-video`, `image-to-video`, `reference-to-video` | | `multi_prompt` | object\[] | no | — | 1–8 items; each `{ prompt, duration_seconds }` | | `prompt` | string | no | — | ≥ 1 character; ≤ 2,500 characters | | `quote_id` | string | no | — | UUID | | `quote_only` | boolean | no | `false` | — | | `references` | string\[] | no | — | ≤ 10 items; only in `reference-to-video` | | `resolution` | enum | no | `1080p` | `720p`, `1080p` | | `scene_id` | string | no | — | UUID | | `video_urls` | string\[] | no | — | 1–5 items; only in `reference-to-video` | | `wait_timeout_seconds` | integer | no | `0` | 0–60 | | `webhook_url` | string | no | — | https URL; ≤ 4,096 characters | The schema is strict. A field that is not in this table gets `400 unsupported_option`. The field names are the roasOS role names. One role has one name in every family. ## Media This endpoint accepts these media roles: * `avatar_id` — identity; the saved character reference sheet, with a derived portrait only where the model requires a single face or source frame * `elements` — identity; one frontal image plus more views, per person or object * `references` — style and appearance references * `image_url` — first frame (or the single source image) * `end_image_url` — last frame * `scene_id` — first frame — a scene this workspace owns * `video_urls` — reference videos | Role | Items | Maximum size | Content types | Duration | | --------------- | ----- | ------------ | --------------------------------------- | -------- | | `end_image_url` | ≤ 1 | 10 MB | `image/jpeg`, `image/png`, `image/webp` | — | | `image_url` | ≤ 1 | 10 MB | `image/jpeg`, `image/png`, `image/webp` | — | | `references` | ≤ 10 | 10 MB | `image/jpeg`, `image/png`, `image/webp` | — | | `video_urls` | ≤ 5 | 100 MB | `video/mp4` | 3–30 s | Every other role is refused: `product_id`, `product_images`, `reference_images`, `scene_image`, `video_url`, `audio_urls`, `audio_url`, `voice_id`. A refused role gets `400 unsupported_option`. Each media URL must be an immutable object on `https://cdn.roasos.com` that your workspace owns. Any other URL gets `400 reference_media_not_workspace_owned` . Refer to [Reference media and uploads](/concepts/reference-media) to learn how to put a file there. ## Prompt Limit 2,500 characters for `prompt`, and 512 characters for each `multi_prompt` shot. * Think in shots: "Shot 1 (3s): …". Define each subject and the setting in shot 1, and keep them the same. * In each shot, write the subject, the action and its motion path, the setting, the camera, and the mood. Use one action and one camera move for each shot. * Write motion as physics: the main movement, then the movement of hair, fabric or particles. * Write each spoken line in quotes with a speaker label and a voice: \[Mia, warm]: "Okay, this texture." * In image-to-video, write only what changes. The first frame sets the scene. * Use each `@Image`, `@Video` and `@Element` tag once. The platform adds a missing `@Image` or `@Video` tag at the end of a single prompt. You must write each `@Element` tag yourself. * End with "No text overlays, no captions." Add text in post-production. ## Not exposed These catalog entries describe omitted inputs, pinned values and source constraints. They do not imply that the provider accepts every listed field. * `video_urls` — these values are not available: `MOV containers`. * `middle_image_urls` — not accepted by this public endpoint. * `voice_ids` — not accepted by this public endpoint. ## Result The result kind is `video`. Always present: `url`. ```json { "data": { "id": "gen_…", "kind": "video", "model": "kling-4", "status": "completed", "result": { "url": "https://cdn.roasos.com/workspaces/ws_01/result" } } } ``` The files are immutable objects on `https://cdn.roasos.com`. A result URL is a valid media input for your next request. ## Example The three steps are always the same: quote, run, poll. cURL JavaScript Python ```bash # 1. Quote. Nothing is spent. curl -X POST https://api.roasos.com/v1/video/kling-4 \ -H "Authorization: Bearer roasos_sk_live_..." \ -H "Idempotency-Key: 0f1c9e2a-kling-4-0001" \ -H "Content-Type: application/json" \ -d '{"quote_only":true,"prompt":"A woman holds the bottle at eye level in a bright kitchen, morning light, static camera."}' # 200 { "data": { "kind": "quote", "quote_id": "q_7Hk...", "credits": 9, "expires_at": "...", # "confirmation": { "model": "kling-4", "example": "3s_1080p" } } } # 2. Run. Same body, same Idempotency-Key, quote_id in place of quote_only. curl -X POST https://api.roasos.com/v1/video/kling-4 \ -H "Authorization: Bearer roasos_sk_live_..." \ -H "Idempotency-Key: 0f1c9e2a-kling-4-0001" \ -H "Content-Type: application/json" \ -d '{"quote_id":"q_7Hk...","prompt":"A woman holds the bottle at eye level in a bright kitchen, morning light, static camera."}' # 201 { "data": { "id": "gen_...", "status": "queued", "credits_reserved": 9 } } # 3. Poll, then read the result. curl https://api.roasos.com/v1/jobs/gen_.../status -H "Authorization: Bearer roasos_sk_live_..." curl https://api.roasos.com/v1/jobs/gen_.../result -H "Authorization: Bearer roasos_sk_live_..." ``` ```javascript const key = process.env.ROASOS_API_KEY; const url = "https://api.roasos.com/v1/video/kling-4"; const idempotencyKey = "0f1c9e2a-kling-4-0001"; const input = { "prompt": "A woman holds the bottle at eye level in a bright kitchen, morning light, static camera." }; const post = (payload) => fetch(url, { method: "POST", headers: { Authorization: `Bearer ${key}`, "Idempotency-Key": idempotencyKey, "Content-Type": "application/json", }, body: JSON.stringify(payload), }).then((response) => response.json()); // 1. Quote. Nothing is spent. const quote = await post({ quote_only: true, ...input }); console.log(quote.data.credits); // 2. Run. const job = await post({ quote_id: quote.data.quote_id, ...input }); // 3. Poll, then read the result. let status = "queued"; while (status !== "completed" && status !== "failed" && status !== "canceled") { await new Promise((done) => setTimeout(done, 3000)); const poll = await fetch(`${url.split("/v1")[0]}/v1/jobs/${job.data.id}/status`, { headers: { Authorization: `Bearer ${key}` }, }).then((response) => response.json()); status = poll.data.status; } const result = await fetch(`${url.split("/v1")[0]}/v1/jobs/${job.data.id}/result`, { headers: { Authorization: `Bearer ${key}` }, }).then((response) => response.json()); console.log(result.data.result); ``` ```python import os, time, requests key = os.environ["ROASOS_API_KEY"] url = "https://api.roasos.com/v1/video/kling-4" headers = { "Authorization": f"Bearer {key}", "Idempotency-Key": "0f1c9e2a-kling-4-0001", } payload = { "prompt": "A woman holds the bottle at eye level in a bright kitchen, morning light, static camera." } # 1. Quote. Nothing is spent. quote = requests.post(url, headers=headers, json={"quote_only": True, **payload}).json() print(quote["data"]["credits"]) # 2. Run. job = requests.post(url, headers=headers, json={"quote_id": quote["data"]["quote_id"], **payload}).json() job_id = job["data"]["id"] # 3. Poll, then read the result. jobs = "https://api.roasos.com/v1/jobs" status = "queued" while status not in ("completed", "failed", "canceled"): time.sleep(3) status = requests.get(f"{jobs}/{job_id}/status", headers=headers).json()["data"]["status"] result = requests.get(f"{jobs}/{job_id}/result", headers=headers).json() print(result["data"]["result"]) ``` ## See also * [`llms.txt` for this model](https://docs.roasos.com/models/video/kling-4/llms.txt) — the same facts as one plain-text file for an agent. * `GET https://api.roasos.com/v1/catalog/kling-4` — the machine-readable form: enums, defaults, limits and prices. Refer to the [API reference](/api-reference/overview). * [Jobs](/concepts/jobs) — how to poll a job and read its result. * Provider endpoints: `kling/v4/image-to-video` · `kling/v4/reference-to-video` · `kling/v4/text-to-video`.