Kling for ads: a practical guide

What Kling does well for ad work

Kling is Kuaishou's AI video and image studio, sold as a consumer app and as an API (Kling AI). For ad work its strengths are lip sync on every API model, native audio, and tools for keeping a person, product or voice consistent across shots.

Model versions differ between the app and the API. Kling 4.0 was announced on 30 Sep 2026: "Kling 4.0 is entering early access, with a wider rollout coming in October" (Kling 4.0 announcement). The API's model list, when checked, still topped out at Kling 3.0, 3.0 Omni and 3.0 Turbo (API capability map).

The two specs that matter most for direct-response ads are lip sync on all models and saved elements. Lip sync covers every talking shot; elements let you reuse the same presenter or product across dozens of versions.

Which ad formats Kling suits

Kling suits talking formats and character-led animation, because lip sync and voice control are published across its models. The table is guidance from published capabilities, not test results.

For a vertical paid-social ad, stay on 9:16. The API aspect ratio list is 16:9, 9:16 and 1:1, so a 4:5 feed version needs a crop in the editor.

Workflow: from script to finished Kling ad

Build elements first, generate shot by shot with lip sync, then assemble and caption in an editor. A 30 to 45 second ad on Kling 3.0 is usually three to six clips of 5 to 15 seconds.

1. Break the script into shots. One spoken line per shot, with a duration under 15 seconds for 3.0 models or under 30 seconds for 4.0.
2. Create elements. Make a presenter image set (front, three-quarter, profile) and a product image set, and save them as elements. Check your plan's element creation quantity before you build a large cast.
3. Fix the voice. Choose one voice per presenter through voice control on 2.6 or a subject voice on 4.0, or generate the voice track separately and lip sync to it. Keep the same voice across every version of the script.
4. Generate each shot. Use image-to-video with the element references. For product shots, set first and last frames so the label lands where your text overlay will sit.
5. Review every take. Check mouth shapes on hard consonants, hand and finger count, label text, and whether the product changes shape between frames.
6. Edit. Cut the shots together, add captions, music and the end card. Put prices, offers and claims in as editor text, not in the generation.
7. Export versions. Swap one variable per version (hook line, presenter element, setting) and keep the rest fixed.

On 4.0, try the hook and first product beat as one 30-second generation, then cut away to inserts. That reduces the number of joins where a face or voice can drift.

Kling prompt structure for ads

Write Kling prompts in a fixed order: who (element), where, action, camera, spoken line, sound. In image-to-video, describe motion and camera rather than appearance, because the image already sets the look.

The templates below are written for this guide. Square-bracket items are placeholders.

UGC selfie hook (9:16, presenter element, lip sync)

[presenter element] holds the phone at arm's length, front camera, vertical.
Setting: [car passenger seat / bedroom / bathroom mirror], daylight.
She raises her eyebrows, pauses, then says to camera:
"[hook line]"
Handheld micro-shake, natural blinking, no background music.

Product hero shot (first and last frame)

Start frame: [product image] lying on [surface], top-down.
End frame: [product image] standing upright, label to camera, centred.
Camera rises and tilts to eye level as the product turns.
[texture detail, e.g. powder settles / light catches the cap].
Sound: soft whoosh, then a single click. No voice.

Podcast two-shot (two subjects with voices)

Podcast studio, two people at mics with over-ear headphones.
Subject 1: [host A element], voice [voice A]. Subject 2: [host B element], voice [voice B].
Static medium two-shot, foam panels and a lamp behind them.
Subject 1 asks: "[question]"
Subject 2 answers: "[answer, one or two sentences]"
Listeners nod while the other speaks.

Animated explainer scene (mascot)

3D animated [mascot element], a [product or object] with eyes and small arms.
Setting: [inside a kitchen cupboard / on a bathroom shelf], toy-like scale.
The mascot hops to the edge of the shelf, turns to camera and says:
"[one-line problem], so I [one-line benefit]."
Camera: slow push-in. Warm, soft lighting, [palette].

Brand-film establishing shot (4.0, 21:9)

Ultrawide establishing shot, 21:9.
[location] at [time of day], [weather and atmosphere].
[one moving element: a train, a figure walking, waves].
Slow lateral track to the right, shallow haze, film grain.
Ambient sound only: [sound bed].

Limits, terms and policy notes

Kling's main limit for ad work is the plan you are on: free output cannot be used commercially without permission. Check the plan before you deliver anything to a brand.

Commercial use. Kling's Terms of Service (4.6) say that "without our written permission, you may not use, reproduce, distribute, and create derivative works of, and make modifications to, the Output for any commercial purposes" (Kling Terms of Service). The Terms of Paid Service say: "KLING AI members' use of the Output for commercial purposes is not restricted", except "for the purposes of developing or offering competitive products or services of KLING AI" (Kling Terms of Paid Service).

On ownership, the Terms of Service (4.4) state: "We do not claim ownership of the Content." Section 4.5 says that without permission, outputs must carry a Kling AI label. The pricing page marks "Generated content is for commercial use" on Standard, Pro, Premier and Ultra, not on Basic (membership plans).

Pricing, as published (USD, monthly toggle).

The first-month price is a "First-purchase offer available once." The page also offers "Yearly 34% Off", "Monthly 12% Off", "One-Time" and "Business" toggles (membership plans). API pricing sits on a separate page that was not captured, so check it directly if you work through the API.

Model availability. Kling 4.0 was in early access in the app when checked, and its 2-minute extension was listed as coming soon. Plan production on the API 3.0 models unless you have 4.0 access.

Platform labels. Meta labels ad images "created or edited using third-party AI tools with AI info" (Meta). TikTok requires the AIGC label or your own clear disclaimer on AI-generated ad content, and its ad policy lists "Making the primary subject say something they didn't actually say, using AI voice-cloning" as significant AI modification (TikTok ads policy). Lip-synced presenters fall in that territory, so label them. More detail: AI-generated content labels on Meta and TikTok ads.

Voices and faces. Do not clone a real person's voice or face without written permission, and never a public figure's. Build presenters from generated references.

Kling in a one-script, every-format workflow

Kling fits the one-script method because saved elements and voices carry over between formats. The same presenter element can read the script as a selfie, then appear as one host in a podcast two-shot, while a mascot element reads the same lines as an animated version.

Keep the script and voice track fixed, and change only the format's picture. The step-by-step method is in one winning script, made in every ad format.

Anglr pays AI creators, editors and agencies on performance to remake winning ad scripts in every format. Anglr provides the winning scripts and angles, and every video gets a QC and compliance review before it goes live. If you already make full ads in Kling and know they perform, see how it works for creators.