For paid ads, this is the mode that matters. Text prompts invent products. Image-to-video animates the one you already shot.

Why use image-to-video instead of text-to-video for ads?

Brand accuracy. In text mode, Kling 4.0 will produce a plausible but wrong bottle, logo or color. In image mode, the model starts from your actual asset, so the label stays recognizable.

The trade-off is motion range. If your source image has the product centered with no motion cues, you get a subtle push-in or a slight rotation, not a dynamic sequence.

How do you prompt motion on a still image?

Do not re-describe the product. Describe what changes: camera path, light shift, or a small physical action.

For a coffee bag: “Slow dolly-in, steam rising from the cup beside the bag, warm side light, shallow depth of field.” The product stays put; the scene breathes.

1. Shoot or source a clean, well-lit image with the product unobstructed.
2. Upload it to Kling and select Image to Video.
3. Enter only motion and camera instructions in the prompt.
4. Choose 5 seconds and standard quality for the first pass.
5. Render, review, then try a Pro-quality pass only if the first is clean.

What resolution and aspect ratios can you get?

Output reaches 1080p on higher tiers. Standard 16:9, 9:16 and 1:1 are supported, which covers YouTube, Reels and feed placements. Square is the least stable for motion and skews slightly during dolly moves.

For paid social, render vertical natively rather than cropping a horizontal clip — cropping cuts resolution and invents artifacts at the new edges.

How many takes should you budget?

Plan for three renders per final shot. One to check motion direction, one to fix lighting, one to deliver. At 10-20 credits each, a 5-shot ad is roughly 150-300 credits, so two entry packs.

Where does it break for product work?

Text on packaging. If your label has fine print, an animated close-up will smear it. Keep the product at medium distance, or add the caption in your editor afterwards.

Reflective surfaces like glass and chrome also flicker frame to frame. Shoot matte, or accept the shimmer as a stylistic choice.

Key Takeaways

• Image-to-video preserves the actual product; text-to-video invents a lookalike.
• Prompt only motion, camera and lighting when using a source image.
• Budget about three renders per final shot, 150-300 credits for a 5-shot ad.
• Fine print on packaging smears; reflective surfaces flicker.

Last updated: September 2026.

Text-to-video has gotten good enough to be useful for B-roll and social cuts. It is not good enough to replace a shoot, and knowing the failure modes saves a lot of wasted credits.

What is Kling 4.0?

It is a video generation model from Kuaishou that produces short clips. V4 improved hand articulation and physical plausibility — v3 still melted fingers during fast gestures.

It accepts text prompts, a single reference image, or image-plus-motion control, and outputs up to 1080p depending on tier.

How do you write a prompt that produces usable motion?

Describe camera, subject, action and lighting in that order. Specificity matters more than length.

Weak: “a person walking.” Better: “Medium shot, woman in a beige coat walking toward camera through a rainy Tokyo street at dusk, neon reflections, handheld camera.” The second gives the model a lens choice and a light source.

1. Pick text-to-video or image-to-video; image mode holds the subject far more consistently.
2. State one primary action. Two simultaneous actions usually produce mush.
3. Set duration — 5 seconds is the reliable sweet spot, 10 seconds costs more and risks drift.
4. Generate, then use Extend only if the clip is clean.

How much does Kling 4.0 cost?

Credits, not dollars per clip, are how it bills. A standard 5-second generation is roughly 10-20 credits depending on quality mode. The entry pack is around $10 for 66 credits, so a handful of real clips per pack once retries are counted.

Video is expensive compared to image generation. Budget two to three attempts for any shot you actually care about.

Where does it still fail?

Text in frame is unreliable — signage and logos dissolve into gibberish. Complex interactions between two people, like a handshake or a dance, often smear. Long continuous motion past 8 seconds tends to warp the subject’s face.

Fast cuts hide all of this. That is why the tool shines in 3-second social clips and struggles in a continuous dialogue scene.

Who is it for?

Social media editors, indie game trailers and anyone needing filler B-roll without a shoot. It is a poor fit for branded content where a client expects exact product representation, because the model will subtly alter your product’s shape.

Key Takeaways

• Kling 4.0 generates 5-10 second clips; 5 seconds is the most reliable length.
• Prompt order: camera, subject, action, lighting.
• Credits run ~10-20 per 5-second standard clip; entry pack is about $10 for 66 credits.
• Text-in-frame and two-person interactions remain weak points.

Last updated: September 2026.