Use Kling O3 to create AI videos with multimodal references. Combine native audio, character consistency and multi-shot generation to shape scenes and stories.
Open Video WorkspaceKling O3 offers multimodal references, native audio, character consistency and multi-shot video generation.
Text-to-video uses a non-empty scene description. Reusable subjects can be defined by 2–4 images or one character video, with optional subject audio references. Custom multi-shot generation supports up to six shots, each with its own description and duration; native audio can be enabled or disabled. Output supports 3–15 seconds and 720p, 1080p or 4K.
Open workspaceImage-to-video starts from one first-frame image or a first and last frame, alongside a scene description. It supports subject references, audio and up to six custom shots, each with a description and duration. Output supports 3–15 seconds and 720p, 1080p or 4K; outside custom multi-shot mode, the aspect ratio follows the image.
Open workspaceReference generation combines a scene description with supported image and subject references, and can include a reference video. Without a video input, it supports custom shots and native audio. When a reference video is present, generated audio must be off; video-only input uses the source aspect ratio, while video with images supports fixed ratios. Reference-image and subject combinations have different limits for each input mode.
Open workspaceVideo transformation uses a source video and a non-empty scene description, with supported image and subject references. Subject references can use multiple images, character videos and audio. Video-only input uses the source aspect ratio; video with images supports 16:9, 9:16 or 1:1. This workflow has its own audio setting and supports 720p, 1080p and 4K output.
Open workspaceKling 3.0 Omni Video Model
Kling 3.0 Omni understands text, images, videos and reusable elements as generation inputs. Different reference information can be combined in one flexible Kling AI video workflow to guide characters, objects, scenes, movement and visual details.
Kling O3 uses references to help preserve characters, products, objects and scenes across shot changes and viewing angles. Reference understanding supports the retention of key visual details in complex scenes with multiple characters.
Kling V3 Omni generates synchronized audiovisual video. Native audio combines dialogue and scene sounds in the same generation workflow, without separating visual and audio tasks. For reference generation with an input video, generated audio must be off.
Kling 3.0 AI turns structured prompts into multi-shot videos, with shot duration, framing, camera angles, character actions, dialogue and camera movement supporting narrative control and scene transitions. Custom shots are supported by text, image and reference generation without an input video; video transformation uses separate controls.
Kling O3 supports videos up to 15 seconds in text, image and reference generation without an input video. Multi-shot sequences, consistent elements, native audio and flexible duration help form longer narratives; video-input workflows have their own duration rules.
Multi-view images define characters, products, objects and scenes, helping preserve detail and consistency. Short videos can provide character features for new scenes, actions and camera angles. Voice characteristics can be attached to reusable character elements for dialogue and storytelling. Characters, objects, images and videos can be combined within each workflow's supported reference limits.
Kling 3.0 Omni supports character-led videos that retain a character's appearance across scenes and shot changes. Reusable character elements support stories, virtual avatars and branded marketing content.
Kling V3 Omni supports product demonstrations, advertising, marketing campaigns and social content. Product images, scene references, motion prompts and native audio in supported workflows turn creative assets into marketing videos.
Kling Video 3.0 Omni supports narrative video that generates multiple shots in one sequence. Framing, shot duration, dialogue, camera movement and character action provide structured storytelling.
Kling 3.0 AI supports dialogue scenes, virtual avatars and virtual characters. Character references, voice references and native audio in supported generation workflows provide reusable characters with consistent visual and voice features.
Text begins with a prompt; Image begins with one first frame or a first/last pair. Reference combines supported subject, image or video references. Transformation starts with an existing video you want to change. Each has separate media controls.
The Image workflow accepts one first frame or exactly two images ordered first then last. Use JPG, JPEG or PNG up to 50 MB per image, with both dimensions at least 300 px and an aspect ratio between 0.4 and 2.5.
Custom multi-shot mode accepts up to six shot descriptions. Custom planning and intelligent shot planning cannot both be enabled. Select a total duration from 3–15 seconds and check the shot timing before submission.
Without a reference video, image references and multi-image subjects share a combined limit of up to seven, reduced for mixed subject types. With a reference video, up to four image references share the allowed subject slots. Follow the counter for the current combination.
Use one MP4 or MOV clip, 3–15.5 seconds and at most 200 MB. Its dimensions must each be 700–4,553 px, with at most 8,294,400 total pixels, aspect ratio 0.4–2 and frame rate 24–60 fps. An existing creation must meet the same limits.
In the Reference workflow, supplying a video requires audio generation to be off. Text and Image can offer audio generation. Transformation has its own audio setting, so check the actual workflow rather than assuming all O3 inputs behave alike.
O3 offers 720p, 1080p and 4K. Text supports 16:9, 9:16 and 1:1. Image and video workflows may use automatic proportions or explicit ratios depending on the input combination and shot mode. Recheck controls when references change.
Use Kling O3 to create AI videos with multimodal references. Combine native audio, character consistency and multi-shot generation to shape scenes and stories.
Open Video Workspace