AIML: text-to-video; separate image-to-video endpoint
Gateway resolution
720p / 1080p; default 1080p
Gateway duration
3–30 seconds; default 5 seconds
Text-to-video framing
16:9, 9:16, 1:1 or 21:9
Audio reference
Kling advertises stereo audio; no audio toggle in reviewed AIML generation schemas
Serving host
AIML API · owner-confirmed integration
API model or route: kling-video/v4/text-to-video
Capabilities & boundaries
What it supports. What the app exposes.
Provider features and app integration are different layers. A feature in the upstream documentation is not automatically enabled in your workspace.
Model and hosting capabilities
Kling's release note describes text-to-video, image-to-video, first/last-frame guidance and broader reference-led workflows. Its platform feature set is wider than either single AIML generation endpoint.
AIML text-to-video accepts a single prompt or multi_prompt shot descriptions, with a total requested duration of 3–30 seconds. This supports planning a continuous shot or a short sequence without treating cuts as guaranteed.
AIML image-to-video uses kling-video/v4/image-to-video with a required image_url and optional tail_image_url. The output aspect ratio follows the supplied images, rather than the text endpoint's aspect_ratio selector.
Kling advertises stereo audio and multilingual lip-sync. Those provider capabilities do not establish a separate audio control or guaranteed spoken dialogue in EZ Ai Assist.
Kling also documents Omni Reference and video editing. These are distinct workflows with separate inputs; they are not implied by a plain text-to-video request.
Before you choose
Kling's release note lists broader resolution features but describes 10-bit HDR at 4K and 1080p as coming soon. The reviewed AIML generation schemas offer only 720p and 1080p; this guide does not promise 4K or HDR through AIML.
Repeated extension up to two minutes is described as coming soon by Kling. It is not a current 120-second duration setting in the reviewed AIML endpoints.
AIML's prompt limit is 2,500 characters, distinct from Kling's platform-level 8,000-token prompt allowance. Do not apply the larger platform limit to the gateway request.
Kling 4.0 Flash is a separate variant with different limits. This page describes Kling 4.0, not Flash or the video-editing endpoint.
Inspect anatomy, foot contact, continuity, shadows and sound before publishing. A prompt describes intent; reproducing the example exactly is not guaranteed.
Only upload media you have rights to use. Obtain appropriate consent for identifiable people and follow provider safety and usage terms.
AIML API controls & integration reference
Text-to-video: prompt, duration, resolution and aspect_ratio are documented fields. Defaults are 5 seconds, 1080p and 16:9. These are gateway defaults, not verified application defaults.
Use prompt or multi_prompt, not both. Each multi_prompt shot has a prompt of up to 512 characters and a duration; the total sequence must fit 3–30 seconds. The single prompt allows up to 2,500 characters.
Image-to-video accepts URL or base64 image inputs, up to 30 MB, at least 300 pixels per side, with input aspect ratio between 0.4 and 2.5. Supply tail_image_url only when using the image endpoint and a supported last-frame workflow.
AIML uses asynchronous POST /v2/video/generations submission and GET /v2/video/generations?generation_id=… retrieval. Handle queued, generating, completed and error states, and consume the returned video only after successful completion.
The reviewed text/image schemas do not expose an audio toggle, seed or negative-prompt field. Do not invent these fields or assume that a shared app control is forwarded to this model.
The API ID shown here is AIML's documented text-to-video reference. The owner confirms AIML as the serving gateway; this website change does not alter or certify the application's exact endpoint selection or fallback behavior.
Made with Kling 4.0
Kling 4.0 in action: a horse in a sunlit meadow
An owner-supplied example made with Kling 4.0 in EZ Ai Assist. The original file is preserved: approximately 10 seconds, 940 × 940 pixels, 24 fps, with an audio track. These are this file's properties, not a claim about native model resolution or guaranteed output settings. The exact prompt is the first copyable example below.
Original examples from EZ Ai Assist. Choose supported controls separately. Some workflows require source media; a text brief does not supply those inputs.
Workflow 01
The horse video: exact original prompt
Copy the prompt used for the supplied example. Select supported settings separately; the original request settings were not supplied, and another generation may differ.
A cinematic video of a majestic chestnut horse galloping at full speed through a sun-drenched wildflower meadow. Golden hour sunlight filters through the distant lush forest, casting long shadows across the swaying grass. The camera tracks the horse in a fluid side-profile shot with a shallow depth of field, highlighting the powerful muscle movement and flowing mane. The atmosphere is vibrant, warm, and serene, captured with a 35mm lens in a high-definition realistic style.
Workflow 02
A controlled product reveal
A simpler shot for evaluating material detail and camera consistency. This is a suggested brief, not an additional generated sample.
A single unbranded brushed-aluminum desk lamp stands on a dark walnut table beside a softly lit window. The camera makes a slow, shallow arc from front-left to front-center while the lamp remains completely still. Preserve its proportions, joint positions and reflections. Warm late-afternoon light, restrained cinematic contrast, one continuous shot, no lettering and no additional objects.
Workflow 03
An atmospheric establishing shot
Keep motion and composition readable before introducing multiple characters or cuts. Review geometry and continuity in the returned clip.
A narrow wooden footbridge crosses a quiet stream in a misty green forest at sunrise. The camera glides slowly forward toward the bridge, staying level above the near bank. Water moves gently beneath intact wooden planks and a few leaves stir in the breeze. Soft golden light enters from the left. Maintain consistent bridge geometry and natural scale in one continuous shot, with no people, titles or abrupt focus changes.
Developer reference
Provider & hosting API pricing
Provider and gateway API rates are not EZ Ai Assist subscription prices or customer charges. Check the application's current quote and plan before generating.
A current model-specific AIML generation rate was not verified in the reviewed text-to-video and image-to-video documentation. No price is inferred from editing, Flash, older Kling models or a sample API usage response.
Pricing reference checked October 24, 2026. Confirm the exact endpoint, resolution and duration with AIML before estimating developer costs.
The supplied sample's generation cost and original request settings were not provided. This website work does not change billing or perform paid generation.
Common questions
Kling 4.0 FAQs
What should I use Kling 4.0 for?
Start with cinematic motion, product reveals and short scene studies. Describe the subject, action, camera and light clearly. Use a supported image-to-video workflow when the initial composition needs a stronger visual anchor.
Can I copy the horse prompt?
Yes. The first example reproduces the owner's prompt exactly, and its Copy button copies that text. The supplied video is one result, not a promise of identical motion, lighting, length or framing on another generation.
Does EZ Ai Assist use AIML for Kling 4.0?
Yes, the owner confirms AIML API as the integration. The documented gateway text endpoint uses kling-video/v4/text-to-video. This guide does not independently certify application controls, request defaults, fallback behavior or subscription access.
How long can an AIML generation be?
The reviewed text-to-video and image-to-video documentation specifies 3–30-second requests with a five-second default. For multi_prompt, the combined shot durations must fit that total. Available application controls must be checked separately.
Can I generate 4K HDR or extend a clip to two minutes?
Not on the basis of these AIML schemas: they expose 720p and 1080p. Kling's release note describes 10-bit HDR and repeated extension up to two minutes as coming soon. Those announcements are not current gateway controls.
How do first and last frames work?
AIML's separate image-to-video endpoint requires image_url and optionally accepts tail_image_url. The images guide the beginning and end, subject to the model's behavior. A text description alone is not a structured image input.
Is audio guaranteed, and can I turn it on or off?
Kling advertises stereo audio, but the reviewed AIML text/image schemas do not expose a separate audio toggle. The supplied example contains an audio track; that does not establish an application control or guarantee particular audio for every request.
What does a generation cost?
A current AIML generation-specific rate was not independently verified. API reference prices and app customer charges are different. Check the current application quote rather than using a neighboring model's price or an editing rate.
Sources & freshness
Check the original references.
Reviewed . Specifications are based on the linked Kling release note and AIML model-specific API documentation. AIML integration and the example's model attribution are owner-confirmed; application handlers and generation settings were not independently traced. Unverified prices and controls are not inferred from older Kling models.