#image-generation #video-editing #video-generation #分镜

Kling AI

Official Kling AI Skill. Call Kling AI for video generation, image generation, subject management, and account quota inquiry. Use subcommand video / image / element / account by user intent. Use when the user mentions "Kling", "可灵", "文生视频", "图生视频", "参考视频", "视频编辑", "文生图", "图生图", "AI 画图", "视频生成", "图片生

Kling AI@klingai-dev

Install

openclaw skills install @klingai-dev/klingai

Language: Respond in the user's language (detect from their message). Use it for explanations, confirmations, errors, and follow-ups. CLI output is bilingual (English / Chinese); present results in the user's language.

Kling AI

Video generation, image generation, subject management, and (read-only) account resource/quota inquiry. Invoke with subcommand video | image | element | account by user intent. Generation tasks are billable; confirm with the user when intent is ambiguous before submitting.

Invocation

From repository root:

bash

node skills/klingai/scripts/kling.mjs <video|image|element|account> [options]

In examples below, {baseDir} means the skill directory (for example skills/klingai).

Routing priority (OpenClaw)

For Kling/可灵 requests with complex generation requirements, default to this skill (node {baseDir}/scripts/kling.mjs ...).
Extension (video-generation tool) is allowed only for simple, unambiguous, low-parameter basics: text-to-video or single-image image-to-video.
Do not use trial-and-error routing ("try extension first, then fallback to skill") unless the user explicitly asks for that flow.

Intent routing (required)

Choose subcommand from user intent first. HTTP API path and default model_name are determined by Route & model.

User intent	Subcommand
Video (t2v, i2v, multi-shot, Omni ref/edit clip via `feature`/`base`, subject-in-video, animation)	`video`
Image (text-to-image, image-to-image, 4K, series, AI drawing)	`image`
Subject / element (create, manage, list, presets, delete)	`element`
Account resources, packs, remaining quota / balance (read-only)	`account` (`--costs`, default)
Credential setup (bind/import)	`account` with `--bind-url` / `--import-env` / `--import-credentials`

Selection rules:

Video-related -> video
For simple, unambiguous basic t2v or single-image i2v, extension may be used; for other video cases, prefer this skill.
Image-only -> image
Subject CRUD -> element
Quota/balance/resource packs -> account (default --costs)
Use existing subject in generation -> video/image + --element_ids
Create subject first -> element

Force skill conditions (any hit -> use this skill):

multi-image input (>=2 images)
Omni/frame control (first_frame/last_frame/image_types)
reference video (--video + feature/base) or video editing
subject/element reuse (--element_ids)
storyboard/multi-shot (--multi_shot, 分镜)
image series (--result_type series/--series_amount, 组图)
extension parameter gaps or ambiguous/unclear parameter intent

Model name strict rule:

Video --model canonical names: kling-3.0, kling-3.0-turbo, kling-2.6, kling-2.5-turbo, kling-3.0-omni, kling-o1.
Image --model canonical names: kling-v3, kling-v2-1, kling-v3-omni, kling-image-o1.
Do not pass aliases as CLI values.
Alias disambiguation: 视频O3 -> kling-3.0-omni; 图片O3 -> kling-v3-omni; only o1/omni1 map to O1 models by intent.

When ambiguous (for example video vs image, or 3.0-omni vs o1), ask user first, then submit.

Preflight checklist (mandatory before submit)

Before any billable submit, pass all checks below. If any check fails, stop and ask user or run --help.

Subcommand is confirmed: video / image / element / account.
Route is confirmed by flags: basic vs Omni (from Route & model).
--model is canonical (no alias values like o3, omni3).
All params come from this SKILL.md or subcommand --help; no undocumented flags.
No conflicting combinations (for example --video_refer_type base + first/last frame, --video feature + --audio native/original, --audio original without --video).
Query mode and submit mode are not mixed.

Anti-fabrication policy (no guessing)

Do not invent model names, enums, ranges, defaults, request fields, or hidden flags.
Do not infer unsupported values from older/other skills.
If value is uncertain, verify with node {baseDir}/scripts/kling.mjs <subcommand> --help.
If user intent is uncertain, ask first; do not submit trial jobs.
If user uses alias words, map to canonical names and pass canonical only.

Cost and submission rules

Every submit is charged; do not submit speculatively.
Confirm intent first when unclear.
On timeout/failure/unexpected result, ask user whether to wait or retry.
Do not auto-resubmit or silently change intent/parameters.

Agent loop & results

Entry: only node {baseDir}/scripts/kling.mjs with video/image/element/account.
Default flow: submit -> poll (~10s interval) -> download to --output_dir.
Keep user updated on long runs (video: submitted -> processing -> succeeded/failed; image/element: succeed/failed).
--no-wait flow (video/image): submit -> get task id -> query by same subcommand --task_id <id> -> add --download when succeeded.
Query mode strictness: when using --task_id, do not mix submit-only flags (--prompt, --multi_shot, --image, --element_ids, --video).
Never print secrets (KLING_API_KEY, api_key).

Presenting results:

Always return task id + local path(s).
If stdout includes an URL, include markdown link as fallback.

Prerequisites

Runtime: Node.js 18+, no extra packages.
Credential priority: KLING_API_KEY (session Bearer as-is) -> stored api_key in .credentials (same Bearer; no JWT). KLING_TOKEN is not recognized.
Official business APIs: Authorization: Bearer <apiKey>. Device bind is unchanged Device Flow; POST /auth/skill/exchange ISSUED returns credential.apiKey. See 接口鉴权.
KLING_API_KEY is never read from env files. --import-env copies the process env value into credentials. --bind-url / --import-credentials / --configure write credentials, not the env var.
Permission/auth errors: report the API message. Rebind only after user confirmation when the stored API Key itself is invalid. Legacy AK/SK files are ignored — ask the user to rebind.
Storage root: default ~/.config/kling, optional KLING_STORAGE_ROOT.
No KLING_API_KEY and no stored api_key: CLI auto-starts bind flow (writes API Key).
account --bind-url: init -> print URL (manual open) -> poll exchange -> persist api_key.
Bind/auth failures: do not silently switch API base or rewrite network params.
Forced rebind (requires user confirmation):
- node {baseDir}/scripts/kling.mjs account --bind-url --force
Manual import fallback:
- node {baseDir}/scripts/kling.mjs account --import-env
- node {baseDir}/scripts/kling.mjs account --import-credentials --api_key "<API_KEY>"
- node {baseDir}/scripts/kling.mjs account --configure
Mask secret values in user-facing text.
Optional behavior (API base, media roots): check subcommand --help.

Quick start

bash

# Show help
node {baseDir}/scripts/kling.mjs --help

# Video
node {baseDir}/scripts/kling.mjs video --prompt "A cat running on the grass" --output_dir ./output
node {baseDir}/scripts/kling.mjs video --image ./photo.jpg --prompt "Wind blowing hair"
node {baseDir}/scripts/kling.mjs video --prompt "Match motion of <<<video_1>>>" --video "https://..." --video_refer_type feature
node {baseDir}/scripts/kling.mjs video --prompt "Change background to ..." --video "https://..." --video_refer_type base
node {baseDir}/scripts/kling.mjs video --multi_shot true --prompt "镜头 1, 5, Sunrise over the hills; 镜头 2, 5, Conflict in the valley;"
node {baseDir}/scripts/kling.mjs video --prompt "A story in three beats: arrival, conflict, resolution"

# Image
node {baseDir}/scripts/kling.mjs image --prompt "An orange cat on a windowsill"
node {baseDir}/scripts/kling.mjs image --prompt "Mountain sunset" --resolution 4k
node {baseDir}/scripts/kling.mjs image --model kling-image-o1 --prompt "<<<element_1>>> on the beach" --element_ids 123456
node {baseDir}/scripts/kling.mjs image --prompt "Character on the beach" --element_ids 123456

# Subject / element
node {baseDir}/scripts/kling.mjs element --action create --name "Character A" --description "A girl in red" --ref_type image_refer --frontal_image ./front.jpg --refer_images ./side.jpg
node {baseDir}/scripts/kling.mjs element --action list
node {baseDir}/scripts/kling.mjs element --action query --task_id <id>

# Account
node {baseDir}/scripts/kling.mjs account --help
node {baseDir}/scripts/kling.mjs account
node {baseDir}/scripts/kling.mjs account --days 90
node {baseDir}/scripts/kling.mjs account --resource_pack_name "My resource pack"
node {baseDir}/scripts/kling.mjs account --bind-url
node {baseDir}/scripts/kling.mjs account --bind-url --force
node {baseDir}/scripts/kling.mjs account --import-env
node {baseDir}/scripts/kling.mjs account --import-credentials --api_key "<API_KEY>"
node {baseDir}/scripts/kling.mjs account --configure

# Query existing task
node {baseDir}/scripts/kling.mjs video --task_id <id> --download
node {baseDir}/scripts/kling.mjs image --task_id <id> --download

Core parameters by subcommand

Do not invent values/ranges/enums/defaults. If unsure, check: node {baseDir}/scripts/kling.mjs <subcommand> --help

video (video generation)

Official docs: 文生视频 3.0, 图生视频 3.0, Omni 视频, 能力地图.

Parameter	Description	Default
`--prompt`	T2V/Omni required. Multi-shot format: `镜头 n, m, words;` (half-width separators; n=index 1–6, m=shot seconds ≥1 summing to `--duration`, words ≤512). Omni refs: `@image_1` `@element_1` `@video_1`.	—
`--image`	Basic i2v: single first-frame image. Omni: comma-separated list. With `--aspect_ratio`, route to Omni.	—
`--image_types`	Omni only. Per-image type aligned with `--image`: `first_frame` / `last_frame` / `refer_image` / empty (`refer_image`). Do not combine `last_frame` with `--image_tail`.	—
`--duration`	Seconds. `kling-3.0` / `kling-3.0-turbo` / `kling-3.0-omni`: 3–15. `kling-o1`: 3–10. `kling-2.6` / `kling-2.5-turbo`: `5` or `10`.	5
`--model`	Path model; see Route & model.	route default
`--resolution`	`720p` / `1080p` / `4k`. `4k` only on `kling-3.0` and `kling-3.0-omni`.	official `720p` if omitted
`--aspect_ratio`	`16:9` / `9:16` / `1:1`. T2V; Omni when no first_frame and not base video edit. Feature video without first_frame: default `16:9`. Do not pass with first_frame or `--video_refer_type base`. I2V has no this field (with `--image` it routes to Omni).	16:9 when sent
`--audio`	`native` / `off` / `original`. 3.0 T2V/I2V and 2.6: `native\|off`. Omni 3.0: `native\|original\|off`. O1: `original\|off`. Turbo/2.5: not sent. `original` requires `--video` (keep source audio). Feature video: `off` only. Base video: not `native`. 2.6 native requires `1080p`.	official `off` if omitted
`--image_tail`	Last-frame image (`contents[].type=last_frame`). Requires `--image`. Do not combine with `--image_types last_frame`. Turbo I2V does not support last_frame.	—
`--element_ids`	Subject IDs (comma-separated). Single `--image` + `kling-3.0` I2V: `contents` `type=element` (max 3). Otherwise Omni.	—
`--video`	Omni reference clip: public http(s) URL only.	—
`--video_refer_type`	`feature` → `feature_video` / `base` → `base_video`.	base
`--multi_shot`	`true` / `false`. Request field only on `kling-3.0` and `kling-3.0-omni` (`settings.multi_shot`). Turbo: write `镜头 n, m, words;` in `--prompt`; do not pass `--multi_shot`. `false` disables 镜头-format prompts. Base video edit: if omitted, CLI sends `false` (server default `true` would fail).	official `true` if omitted (3.0 / 3.0-omni), except base edit → `false`
`--output_dir`	Output directory.	`./output`
`--task_id`	Query via `GET /tasks?task_ids=`; pair with `--download`.	—

Do not pass: --mode, --sound, --shot_type, --multi_prompt, --keep_original_sound, --negative_prompt.

Model alias reminder (do not pass aliases to --model):

omni3/omni v3/o3/video o3/视频O3 -> kling-3.0-omni
o1/omni1 -> kling-o1 (video) or kling-image-o1 (image)

Omni contents image rules:

url cannot be empty (URL or Base64).
Types: first_frame / last_frame / refer_image. Last-frame requires first-frame. No last-frame-only. kling-3.0-omni and kling-o1 both support last_frame (o1 has extra limits below).
--image_tail requires --image. Do not pass --image_tail together with --image_types last_frame.
With --video: max 4 images. Without --video: max 7.
kling-o1: when image count > 2, no last_frame. First+last: no extra refer_image.
Frame generation cannot combine with --video_refer_type base.
kling-o1 + feature_video: first_frame only, no last_frame.
kling-o1 first_frame only (no refer_image, no --video): --duration only 5 or 10.

Omni contents element rules:

element_id cannot be empty; id is element_1…
Frame generation supports up to 3 subjects.
First+last frame with kling-o1: subjects unsupported.
With --video: image_count + element_count <= 4; otherwise <= 7.

Omni contents video rules:

Max one URL.
--video_refer_type: feature / base (default base).
feature_video: --audio must be off; do not set --multi_shot false; no first_frame → send aspect_ratio (default 16:9).
base_video: no first/last frame; no --multi_shot true; if --multi_shot omitted, CLI sends false; --audio cannot be native; do not send aspect_ratio.
--audio original requires --video (keep source audio). Feature cannot use original.

Compact examples:

bash

# explicit frame marking by intent
node {baseDir}/scripts/kling.mjs video --model kling-3.0-omni --image a.jpg,b.jpg,c.jpg --image_types first_frame,refer_image,last_frame --prompt "..."

# with reference video: image count <= 4
node {baseDir}/scripts/kling.mjs video --video "https://..." --video_refer_type feature --audio off --image a.jpg,b.jpg --prompt "..."

image (image generation)

Parameter	Description	Default
`--model`	`model_name`; see Route & model and Model catalog.	route default
`--prompt`	Image prompt (required). Omni: `<<<image_1>>>` / `<<<element_1>>>`.	—
`--image`	Basic: single image. Omni: image list (comma-separated).	—
`--resolution`	`1k` / `2k` / `4k`; `4k` routes to Omni.	1k
`--aspect_ratio`	`16:9` / `9:16` / `1:1` / `auto` (`auto` Omni only, and only with `--image`).	basic: `16:9`; Omni t2i: `16:9`; Omni with `--image`: `auto`
`--n`	Result count 1–9 (`result_type=single`).	1
`--negative_prompt`	Basic API only.	—
`--result_type`	`single` / `series` (`series` is Omni and i2i-only).	single
`--series_amount`	2–9 or `auto` for `result_type=series`. `auto` = 智能选择数量.	4
`--element_ids`	Subject IDs (comma-separated). `kling-v3` generations: `element_list`. Omni: `element_list`. `kling-v2-1` unsupported.	—
`--output_dir`	Output directory.	`./output`
`--task_id`	Query task id; pair with `--download`.	—

Notes:

n and series_amount apply to different modes.
series is i2i-only, so --result_type series requires --image.
--negative_prompt is generations-only and cannot combine with --image.
Omni aspect_ratio=auto needs --image. Text-to-image Omni (no --image) defaults to 16:9; do not pass auto.

Omni refs rules (image):

image cannot be empty (URL or Base64).
element_id cannot be empty.
image_count + element_count <= 10.

element (subject management)

Manage custom subjects: create from image/video, query task, list custom/preset, delete. Use element_id in video/image with --element_ids for reusable subject consistency.

Parameter	Description
`--action create`	Create subject; requires `--name` (<=20), `--description` (<=100), `--ref_type`
`--ref_type`	`image_refer` (requires `--frontal_image` and `--refer_images`) / `video_refer` (requires `--video`)
`--frontal_image`	Front reference image (`image_refer`)
`--refer_images`	Other reference images (comma-separated, required 1–3). Official API: must differ from the frontal image.
`--video`	Reference video (`video_refer`)
`--voice_id`	Optional voice bind (image_refer: 人物/类人形象)
`--action query --task_id <id>`	Query creation task
`--action list`	List custom subjects
`--action list-presets`	List preset subjects
`--action delete --element_id <id>`	Delete subject

If the user only provides a frontal image: ask whether to (1) add 1–3 other-angle images, or (2) explicitly reuse the frontal file as --refer_images. Do not submit without --refer_images. The CLI does not silently copy --frontal_image into refer_images.

account (resource & quota inquiry, optional credential setup)

Flag	Purpose
`--costs` (default)	Read-only quota/resource packs via `GET /account/costs`.
`--bind-url`	Device bind with polling; prints URL for manual open; optional `--force`.
`--import-env`	Read `KLING_API_KEY` and persist.
`--import-credentials`	Persist `--api_key`.
`--configure`	Interactive API Key input and save credentials.

All bind/account files persist under storage root (~/.config/kling by default, or KLING_STORAGE_ROOT).

--costs query params:

Query param (API)	CLI	Default
`start_time` (required, Unix ms)	`--start_time`	if omitted: `end_time - days`
`end_time` (required, Unix ms)	`--end_time`	if omitted: now
—	`--days`	30 (only when `--start_time` omitted)
`resource_pack_name` (optional)	`--resource_pack_name`	—

Run node {baseDir}/scripts/kling.mjs account --help for details. Run node {baseDir}/scripts/kling.mjs video --help, image --help, or element --help for full params.

Route & model (CLI: `kling.mjs` + flags -> default `model_name`)

Agents call node {baseDir}/scripts/kling.mjs <video|image|element|account> with flags. --model selects the official path/model. Video encodes model in the URL (/text-to-video/kling-3.0); image still sends model_name. Must be exact canonical spelling. If --model is omitted, route defaults apply. CLI guardrails reject incompatible model/route and invalid audio / frame combinations before submit.

Routing decision tree (must follow)

Choose subcommand from intent: video / image / element / account.
Determine route triggers:
- any Omni trigger -> Omni route
- otherwise -> basic route
Validate model-route compatibility:
- Omni route accepts only Omni-capable canonical models
- basic route rejects Omni-only models
Validate strict parameter combos (audio, multi_shot, frame rules, ref limits).
If uncertain, run --help or ask user; never guess-submit.

Video (`video` subcommand)

Omni routing triggers (any of these -> /omni-video/{model}):

--video
comma in --image
--image + --aspect_ratio
--element_ids without a single I2V first-frame (or with turbo/2.6/2.5)
explicit --model kling-3.0-omni or --model kling-o1

Otherwise:

T2V: no --image → POST /text-to-video/{model}
I2V: single --image (optional --image_tail, optional --element_ids on kling-3.0 max 3) → POST /image-to-video/{model}

--multi_shot does not force Omni; it still follows the routing triggers above. Turbo has no settings.multi_shot; use prompt 镜头 n, m, words; only.

Video routing (CLI)	Default if `--model` omitted	Allowed `--model`
T2V / I2V	`kling-3.0`	`kling-3.0`, `kling-3.0-turbo`, `kling-2.6`, `kling-2.5-turbo`
Omni	`kling-3.0-omni`	`kling-3.0-omni` (default), `kling-o1` (explicit)

Image (`image` subcommand)

Omni routing triggers (any of these -> omni-image API route):

explicit --model kling-v3-omni or --model kling-image-o1
--result_type series
--resolution 4k
--aspect_ratio auto
comma in --image

--element_ids alone stays on /v1/images/generations with kling-v3 element_list (kling-v2-1 rejected). 4k/series/multi-image still go Omni.

Else -> basic generations route (text-to-image / image-to-image).

Image routing (CLI)	Default if `--model` omitted	Allowed `--model`
Basic `/v1/images/generations`	`kling-v3`	`kling-v3`, `kling-v2-1`
Omni `/v1/images/omni-image` (4k / series)	`kling-v3-omni`	`kling-v3-omni`, `kling-image-o1`
Omni `/v1/images/omni-image` (other)	`kling-image-o1`	`kling-image-o1` (official default), `kling-v3-omni`

Model catalog (by name)

Common aliases (understanding only; do not pass aliases to --model):

omni3, omni v3, 视频O3, O3, o3 -> kling-3.0-omni (video) / kling-v3-omni (image)
o1, omni1 -> kling-o1 (video) or kling-image-o1 (image)

--model input rule: pass only canonical names from this table.

Model	Valid on	Notes
`kling-3.0`	T2V / I2V	Default video basic. Resolution 720p/1080p/4k; audio native/off; multi_shot. I2V contents may include `element` (max 3).
`kling-3.0-turbo`	T2V / I2V	No audio / `settings.multi_shot`; I2V first_frame only (no last_frame). Multi-shot via prompt `镜头 n, m, words;` only.
`kling-2.6`	T2V / I2V	Duration 5 or 10; audio native/off (native needs 1080p). Last-frame: 1080p silent.
`kling-2.5-turbo`	T2V / I2V	Duration 5 or 10; no audio field. Last-frame: 1080p only.
`kling-3.0-omni`	Omni video	Default Omni video. audio native/original/off (`original` needs `--video`); 4k; first+last frame; feature/base video.
`kling-o1`	Omni video	Duration 3–10; audio original/off; no 4k; no multi_shot field. First-frame-only: 5 or 10s. Feature video: no last_frame. First+last: no extra refs/subjects.
`kling-v3`	Basic image	Default basic image. 1k/2k. Supports `element_list`.
`kling-v2-1`	Basic image	Official generations `model_name`. No subjects.
`kling-v3-omni`	Omni image	1k/2k/4k; series + `auto`. Default Omni when 4k or series. Text-to-image: no `aspect_ratio=auto` (default `16:9`).
`kling-image-o1`	Omni image	Official omni-image default. Capability map: 1k/2k (not 4k, not series). Text-to-image default `16:9`; `auto` only with `--image`.

Principle:

Set task flags first (--image, --element_ids, --video, --multi_shot, ...).
Omit --model to use route defaults.
If --model is explicit, it must match route implied by flags.

When to use Omni; element vs image reference

Which route: follow Route & model triggers. Prefer Omni when you need multi-image composition, 4K/series, reference-video edit, or mixed image+element Omni prompts (<<<image_1>>> / @image_1). Single-image + subject on kling-3.0 I2V or kling-v3 generations stays on the basic route.

Prefer plain image reference for simple tasks. Create element first only when user explicitly wants reusable subject consistency across outputs.

Prompt template syntax (video / image Omni)

In Omni video, pass media/subjects by flags; reference in --prompt with official @ ids (contents[].id):

@image_1 -> first --image (@image_2, ...)
@element_1 -> first --element_ids (@element_2, ...)
@video_1 -> --video clip

Omni image still uses <<<image_1>>> / <<<element_1>>> in --prompt (official /v1/images/omni-image). Those placeholders require the Omni route (explicit --model kling-image-o1 / kling-v3-omni, or 4k/series/multi-image). kling-v3 generations uses element_list without <<<>>>.

Notes

Timing: video ~1–5+ min; image ~20–60 s; subject creation ~30 s–2 min.
Retention: platform may remove assets after ~30 days; save outputs locally.

Reference

Official docs index: https://klingai.com/document-api/llms.txt
Official docs home: https://klingai.com/document-api/
API endpoint quick map in this package: reference.md

Kling AI

Install

Kling AI

Invocation

Routing priority (OpenClaw)

Intent routing (required)

Preflight checklist (mandatory before submit)

Anti-fabrication policy (no guessing)

Cost and submission rules

Agent loop & results

Prerequisites

Quick start

Core parameters by subcommand

video (video generation)

image (image generation)

element (subject management)

account (resource & quota inquiry, optional credential setup)

Route & model (CLI: kling.mjs + flags -> default model_name)

Routing decision tree (must follow)

Video (video subcommand)

Image (image subcommand)

Model catalog (by name)

When to use Omni; element vs image reference

Prompt template syntax (video / image Omni)

Notes

Reference

Related skills

Route & model (CLI: `kling.mjs` + flags -> default `model_name`)

Video (`video` subcommand)

Image (`image` subcommand)