KLING 3.0
The fast path from prompt to shot
AVAILABLE ON BEEBLE
KLING 3.0
Text, image, reference, or video in. Native 4K with sound out.
Live on Beeble today.
THE MODEL
KLING 3.0
AT A GLANCE
RESOLUTION
4K
NATIVE 4K OUTPUT
Pick 720p, 1080p, or native 4K per clip — 2160p straight from the model, no upscaling pass.
DURATION
15s
ANY LENGTH, 3 TO 15
Dial the exact clip length, second by second, from 3 to 15. No snapping to a 5-or-10 grid.
MODES
5
FIVE WAYS IN
Text to video, image to video, reference to video, video edit, and reference video. One family, five doors.
REFERENCES
4
SUBJECTS, LOCKED ON-MODEL
Up to 4 reference images per job on Kling 3 Omni, each addressable right in the prompt as Image 1 to Image 4.
THE FAMILY
PICK YOUR VARIANT
Same durations, same resolutions, same prompts. Omni adds reasoning and takes video.
KLING 3 OMNI
The reasoning variant — plans the shot before rendering it
01/ NATIVE AUDIO
SOUND SHIPS
WITH THE SHOT
Audio is generated in the same pass as the pixels — ambience, effects, the room the shot lives in. It's a per-clip toggle, so switch it off when the cut only needs picture.
02/ KLING 3 OMNI
IT READS THE SCRIPT
BEFORE IT SHOOTS
Kling 3 Omni is the reasoning variant: it decomposes your prompt before generating, so multi-clause shot descriptions and chained camera moves land the way you wrote them.
03/ VIDEO TO VIDEO
EDIT THE CLIP
YOU ALREADY HAVE
Hand Omni an existing clip. Video Edit swaps characters, restages the environment, or changes the style while the source's motion and camera language stay put. Reference Video generates new shots that inherit the source's cinematography.
04/ BEYOND SDR
SDR IN.
16-BIT EXR OUT.
Straight out of the model, the clip is SDR. Run it through SDR to HDR on Beeble and finish with a true 16-bit ACES2065-1 EXR sequence. Blown highlights rebuilt, crushed shadows cleaned, ready for the grade.
FOR STUDIOS
ENTERPRISE READY
Kling on Beeble ships with the same guarantees as every model we host. Built for studios that answer to legal.
FULL COMMERCIAL USE
Everything you generate with Kling on Beeble is yours to ship — films, ads, client work. No separate license to chase.
NO-TRAINING GUARANTEE
Your uploads and generations never train the model. Guaranteed in writing.
FREQUENTLY
ASKED
QUESTIONS
Kuaishou's video generation model. Beeble runs the full family — Kling 3.0 and the reasoning variant Kling 3 Omni — in Canvas and the Video Generator.
720p, 1080p, and native 4K, with any clip length from 3 to 15 seconds. Audio is generated with the shot and can be toggled per clip. Text and reference modes render 16:9, 9:16, or 1:1; image modes follow your start frame.
The reasoning variant of Kling 3.0. It decomposes the prompt before generating, so layered, multi-clause direction is followed more faithfully — and it adds reference-to-video and two video-to-video modes on top of text and image.
Yes, three ways. Image-to-video takes a first and an optional last frame. Reference-to-video takes up to 4 reference images, each addressable right in the prompt. The video-to-video modes take an existing clip — edit it in place, or use it as a cinematography reference for new shots.
Text-to-video, image-to-video, and reference-to-video all render up to native 4K. The two video-to-video modes output 1080p or 720p.
No. Never. Your uploads and generations never train the model, and the no-training guarantee is in writing.
Kling generations are billed in Beeble credits per clip, by the second. Current rates for every model and resolution are in the credit policy.