Kling 4.0 vs Kling 3: What Changed and When to Switch
Kling 4.0 doubles the take and widens the control panel. Kling V3 is still the one you can call today. Here's what actually changed, and which shots belong to which model.
Every new version number makes the old one look tired. That's the trap. Kling 4.0 is out on Kling's own site, the clips people are posting are long and confident, and the obvious question is whether everything you built on Kling 3 is suddenly obsolete.
It isn't. But a few things really did change, and they change how you plan a shot, not just how good it looks. This is a side-by-side of Kling 4.0 vs Kling 3 built only from what Kling has confirmed and what Kling V3 does on each::labs today, so you can decide which shots should wait for the new model and which ones can ship this week.
Where Each Model Stands Today
Kling 4.0 is out on Kling's own app and website.
Kling's official global launch release says API access for enterprise customers and developers is coming "soon", and press reports say the API is opening to them in phases. Kling's own developer pages hadn't confirmed that when we checked, so treat it as a rollout in progress, not a door that's open for everyone. It's coming to each::labs, and you can get ready on Kling V3 today.
Kling V3 is the opposite story. It's a full family you can call right now on each::labs: Standard and Pro text-to-video, image-to-video in both tiers, a native 4K image-to-video endpoint, Turbo variants for fast iteration, and Motion Control. Kling 4.0 isn't on each::labs yet. It's coming, and the rest of this post is about what to do in the meantime.
Clip Length: 15 Seconds vs 30, Plus Extension
This is the headline difference, and it's bigger than it looks. Kling V3 generates clips from 3 to 15 seconds on each::labs. Kling 4.0 generates up to 30 seconds natively in a single generation, and Kling says you can upload an existing video and extend it to a total of up to two minutes.
Doubling the take doesn't just mean fewer cuts. It means the model has to hold a character, a light setup and a camera path for twice as long, and you have to direct twice as long. A 15-second clip is a moment. A 30-second clip is a scene with a beginning, a turn and an ending. If your work is mostly product loops, social hooks and 8-second inserts, the extra length changes very little for you. If you make spots, trailers or narrative pieces where stitching was the painful part, this is the reason to care.
The extension feature deserves a careful read. Two minutes is a total length built on top of a clip, not a two-minute generation. Your first 30 seconds become the anchor, so the opening deserves more attention than it ever did on Kling 3.
References: What You Show the Model
Kling V3 already lets you steer a shot with images. The 4K image-to-video endpoint takes a start image, accepts an optional end image, and matches the aspect ratio of the frame you give it. That's a strong, simple contract: here's where we start, here's where we land.
Kling 4.0 widens that contract a lot. Its Omni Reference system takes up to 15 multimodal references, and the launch release says those can be images, video and audio. On top of that it accepts white-model and depth-map reference videos to guide performance, action, camera movement and pacing. A white model is the grey, untextured figure from a 3D blockout. A depth map is a grayscale video where brightness means distance. Both carry motion and space without carrying a look, which is exactly what you want when the prompt and your style references should decide the look.
Here's our take: fifteen references is not a target. It's a ceiling. The teams that get the most out of this will give each reference one clear job (this is the character, this is the room, this is the voice) rather than dumping a moodboard in and hoping. If you've been using Kling V3 Motion Control to copy a move from a reference clip, you already have the habit. Kling 4.0 extends it from one performance to the whole shot.
Structure: Multi-Shot Prompts vs Ten Keyframes
Kling 3 introduced a real way to build sequences inside one generation. On each::labs, Kling V3 accepts a multi-shot prompt list, so you can describe a sequence beat by beat with consistent physics and lighting across shots. One catch worth knowing: the end image works with a single prompt and not with multi-prompting, so you choose between a fixed landing frame and a multi-shot sequence.
Kling 4.0 moves structure from text into anchors. It supports up to 10 keyframes for scene planning. That's a different mental model. With multi-shot prompts you describe what happens. With keyframes you pin what the frame looks like at specific moments, and the prompt describes the motion between them. For a 30-second take that matters, because ten anchors spread across half a minute give you a checkpoint every few seconds instead of one at each end.
Prompts and Audio: Longer Instructions, More Languages
The prompt window grows to 8,000 tokens in Kling 4.0. Don't read that as permission to write longer adjectives. Read it as room for structure: beats with timings, camera moves in order, dialogue lines with delivery notes, and what each reference is for. Kling V3 already rewards order. The each::labs page for Kling V3 Pro recommends a subject, action, environment, lighting, camera and audio sequence, and that structure carries over well.
Audio is where Kling 3 was already strong. Kling V3 generates native audio with lip-sync on each::labs, and the 4K image-to-video spec lists Chinese and English for speech. Kling's launch release says Kling 4.0 improves multilingual support across English, Japanese, Korean, Spanish, Portuguese, German, French, Hindi and more, and names American, British and Indian English accents. If you localize video for several markets, that list is the most practical upgrade in the release.
What Kling 4.0 Hasn't Shipped Yet
Two features that got a lot of early attention are still marked "soon" in the official release: 10-bit HDR output at 4K and 1080p, and the 21:9 ultrawide format. Kling pitches HDR for backlit, night and neon scenes, where more detail survives in highlights and shadows and the grade has more room. That's genuinely useful, but you can't plan a deliverable around it today.
So if a brief needs 4K right now, Kling V3 is the one that delivers it on each::labs, through its native 4K image-to-video and text-to-video endpoints. Comparing a shipped feature against an announced one is how teams lose a week.
Kling 4.0 vs Kling 3: Which Shots Go Where
Put the two side by side and the split gets clear. Keep these on Kling V3 today: product shots and social clips under 15 seconds, anything that needs native 4K now, start-to-end transitions where you control the landing frame, motion transfer with Motion Control, and fast drafts on the Turbo variants.
Save these for Kling 4.0: single takes longer than 15 seconds, scenes where a character has to stay consistent across a full spot, shots you want to block in 3D first and render later, dialogue in the newer languages and accents, and anything that leans on audio or video references rather than stills.
The smart move is to stop thinking of it as an upgrade and start thinking of it as routing. Our guide to calling Kling, Veo and Sora through one endpoint shows the pattern: pick the model per shot behind one integration, so adding a new one is a config change, not a rebuild.
How to Get Ready for Kling 4.0 on each::labs
Build the pipeline on Kling V3 now and design it for the longer take. Write your prompts in timed beats even when the clip is 10 seconds, because that's the format a 30-second take will want. Keep your references organized by role (character, location, style, voice), since that's how Omni Reference rewards you. And if you use 3D at all, start saving blockouts and depth passes for the shots you care about. They'll be inputs soon.
For more on planning long takes, our Kling 4.0 prompting guide breaks half a minute into beats, and nine things to know about Kling 4.0 covers the launch in detail. When Kling 4.0 lands on each::labs, it'll sit next to the rest of the text-to-video models, and you'll be able to route your longest shots to it without touching the rest of the stack.
Frequently Asked Questions
Is Kling 4.0 better than Kling 3?
Better at different things. It holds a single take twice as long, takes far richer references and adds keyframes, which matters for scenes and spots. For short clips and native 4K delivery today, Kling V3 is still the practical choice.
When can I use Kling 4.0 on each::labs?
It's already available on Kling's own website, and press reports say the developer API is opening in phases. It isn't on each::labs yet, but it's coming.
Can I use Kling 4.0 on each::labs?
It's coming to each::labs, and until it arrives you can run the full Kling V3 family there, from Turbo drafts to native 4K.
How long are Kling 4.0 videos compared with Kling 3?
Kling 4.0 reaches 30 seconds in a single generation, while Kling V3 tops out at 15 on each::labs. Kling 4.0 can also extend an uploaded video to a total of up to two minutes, so the real ceiling is a sequence, not a clip.