Kling 4.0 is promised for October 2026, but only the lighter Kling 4.0 Flash is live, and only for Ultra Yearly subscribers. This article sorts what Kling has confirmed from what is still pending: the release window, beta access, 30 second clips, 4K output, Omni Reference, stereo audio and missing pricing, plus a Kling v3 workflow you can run today.
Kling 4.0 has a launch window, but not a launch day. Kling AI says the full model arrives "this October," and as of October 8, 2026 no date has been published. What you can actually touch right now is Kling 4.0 Flash, a lighter sibling that opened on September 28 to a limited group of Ultra Yearly subscribers. Everything else, from 30 second clips to 4K output and a new reference system, is announced but not yet in public hands. This article separates what is confirmed from what is still a promise, lists every announced feature in one place, and shows what you can run today on PicassoIA while you wait.
💡 Quick answer: Full release: October 2026, no exact date. Beta: Kling 4.0 Flash, Ultra Yearly subscribers only, live since September 28. Pricing and API access: not announced.
AI video launches generate a lot of noise, so it helps to sort claims by source. Everything below comes from Kling's own announcement and release notes, as repeated by third-party trackers during the first week of October 2026. Where a detail is missing, this article says so instead of guessing.
Kling's announcement says, in short, that Kling 4.0 is coming this October. That is the entire commitment. There is no day, no time zone and no region list. A month is a wide target, so build plans that survive a slip into early November, and avoid promising a client a Kling 4.0 deliverable on a fixed date.
Flash is the part that exists today. Early access began on September 28, 2026, and the spec is modest compared with the full model:
Think of Flash as a preview of the generation, not a stand-in for it.
Still unannounced:
The announcement names Ultra Yearly subscribers and no other plan. Monthly, Pro and free accounts are not mentioned, so assume no access until Kling says otherwise. Access is also limited to the Kling web app. Several third-party API routers reported in early October that they do not offer either Kling 4.0 tier, and none had a public model identifier to plug into.
Kling has not published pricing for Kling 4.0 or Flash, and it has not said whether existing credit allowances apply. Without a price per second you cannot budget a campaign, and without an API you cannot automate one. If your workflow depends on either, you are waiting.
💡 Our take: Buying a yearly plan only for Flash is hard to justify. Flash caps at 720p and 20 seconds, so it does not deliver the headline features: 4K, 30 second clips and frame anchors. If you already hold a yearly plan, test it. If you do not, wait for the full release and its price.
These specs describe the full model. Flash has none of them unless the list above says so.
The headline spec is length. The full model generates 3 to 30 seconds natively, in a single pass. For comparison, Kling v3 Video on PicassoIA tops out at 15 seconds. Thirty seconds is long enough for a full ad spot or a story beat without stitching clips together, which is where most continuity errors creep in: faces drift, lighting shifts, props change.
The full model lists 720p, 1080p and 4K, plus a 21:9 ultrawide option for cinematic framing. One caution: the release notes mark 10-bit HDR at 4K and 1080p as coming soon, so do not plan an HDR pipeline for day one. Flash is 8-bit SDR.
A 4K file is still useful for reframing. You can cut a 1080 by 1920 vertical delivery from a 3840 by 2160 source file without losing pixels, which turns one render into a landscape cut and a phone cut.
Kling describes multi-frame control with up to 10 anchor images placed along the timeline. Instead of hoping the model lands on the shot you pictured, you should be able to pin the opening, the middle beats and the ending, and let it fill in the motion between them. Kling v3 Video on PicassoIA gives you two anchors today, a start image and an end image. Ten anchors would let you storyboard a 30 second scene almost shot by shot.
On the full model, Omni Reference accepts up to 15 multi-modal references: still images, short video clips (up to 30 seconds of video in total) and subject elements. Voice references are supported too, though the notes list no music or sound effect clips. The point is consistency, the same character, product and location across every shot.
Kling v3 Omni Video on PicassoIA already does a smaller version of this with up to seven reference images.
Audio gets a visible upgrade on paper: two-channel stereo, more precise lip sync, and spoken output in nine named languages (Chinese, English, Japanese, Korean, Spanish, Portuguese, German, Indonesian and Hindi) plus accents and dialects. On-screen text generation is listed as well.
Editing is the other addition. The notes describe re-editing existing footage with up to five clips of 3 to 30 seconds each, 30 seconds combined. Video extension up to two minutes is marked coming soon. Prompts can run to 8,000 tokens, a far longer window than the 2,500 characters that Kling v3 Video accepts.
| Spec | Kling 4.0 Flash | Kling 4.0 (full) | Kling v3 Video on PicassoIA |
|---|---|---|---|
| Status on Oct 8, 2026 | Live, Ultra Yearly only | Announced for October | Available |
| Length | 3 to 20 seconds | 3 to 30 seconds | Up to 15 seconds |
| Resolution | 720p | 720p, 1080p, 4K, 21:9 option | 720p or 1080p |
| Color depth | 8-bit SDR | 10-bit HDR coming soon | Not listed |
| Audio | Not detailed | Stereo, 9 named languages | Optional native audio |
| Frame control | Not listed | Up to 10 anchors | Start and end image, up to 6 shots |
| References | Omni Reference supported | Up to 15 | Up to 7 images in the Omni version |
Where a cell says "Not detailed" or "Not listed," the sources did not mention it. That is not proof the feature is missing.
Kling 4.0 is not listed on PicassoIA at the time of writing, and no date for it has been announced there. Kling v3 Video and its siblings are available, and they are the closest working rehearsal for a Kling 4.0 workflow: multi-shot scripts, start and end frames, native audio and reference images all exist today.
| Model | Best for |
|---|---|
| Kling v3 Video | Text or image to video, up to 15 seconds, 720p or 1080p, up to 6 shots |
| Kling v3 Omni Video | Reference images (up to 7), a reference video, and a video editing mode |
| Kling v3 Motion Control | Animating a character with motion control |
| Kling v2.6 | Cinematic text to video from the previous generation |
| Kling Avatar v2 | Animating a face into a video |
A three-shot, 15 second script looks like this:
[
{"prompt": "Wide shot of a baker opening a wooden shop door at sunrise, warm light spilling onto a cobblestone street", "duration": 5},
{"prompt": "Close-up of hands dusting flour over fresh dough, steam rising, shallow depth of field", "duration": 5},
{"prompt": "Medium shot, the baker slides a tray of golden loaves onto a counter and smiles at a customer", "duration": 5}
]
Use one formula every time: subject and starting pose, action over time, camera move, light, sound. A working example: "A fisherman in a yellow raincoat stands at the end of a wet pier at dawn, then lifts a net and swings it out over the water. Slow dolly-in at shoulder height, soft overcast light, gulls and gentle waves in the audio."
💡 Rehearse now: Write your 30 second Kling 4.0 shot list today as two or three multi-shot clips on Kling v3 Video. When the full model lands, the same list becomes a single pass, and your prompts are already tested.
New releases tend to bring crowded queues and a lot of trial and error. A little preparation turns the first week into production instead of experiments.
Omni Reference takes up to 15 inputs, so collect them per project before the model arrives. Gather character photos from several angles in consistent light, product shots on a clean background, two or three location stills and, if you plan to use a voice reference, a short clean voice sample. One folder per project, named by role (hero, product, location, voice), saves a lot of hunting later.
A 30 second clip needs structure, or it drifts. Break it into beats and write each one with the prompt formula above:
| Seconds | Beat |
|---|---|
| 0 to 5 | Wide establishing shot |
| 5 to 12 | Subject enters, medium shot |
| 12 to 20 | Close-up on the main action |
| 20 to 26 | Reaction or cutaway |
| 26 to 30 | Final wide shot, hold |
Pricing is unannounced, so do not lock a budget around a guess. Plan for slower queues in launch week, and send drafts through Seedance 2.5 Lite first so paid renders go only to shots you have already approved.
If your deadline cannot wait for October, these PicassoIA models fill the gaps. The descriptions come from each model's own listing.
| Model | Listed strength |
|---|---|
| Seedance 2.5 | 30 second AI videos |
| Seedance 2.5 Lite | Free and unlimited video generation up to 10 seconds |
| Veo 3.1 | 1080p text to video |
| Sora 2 Pro | HD video from prompts |
| Wan 3 | Cinematic video |
A practical split: draft shots on Seedance 2.5 Lite until the motion and framing feel right, then render finals on Kling v3 Video in pro mode. If you need a single long take today, Seedance 2.5 is the one listed for 30 second output. The "free and unlimited" wording comes from the model listing, so check the current plan terms on the page before you rely on it.
Waiting for a launch date is a poor use of a good idea. Pick a scene you would give to Kling 4.0, write it as a three-shot script, and run it through Kling v3 Video this week. Add reference photos with Kling v3 Omni Video if you need the same character across shots. By the time the full model arrives, you will have a tested shot list, prompts that work and a clear picture of what 30 seconds should look like.
Open PicassoIA, browse the video models, and start your first clip today.
