Kling AI

Kling 2.5 Turbo separates generation by input type and processing mode. Text-to-video builds the entire shot from the prompt, while image-to-video treats the uploaded image as the starting visual state and uses the prompt to direct motion and camera behavior.

Kling Avatar v2 animates a single portrait from an uploaded audio track. The audio determines the output duration and drives lip movement, while an optional prompt can guide expression and motion. Standard and Professional modes trade generation cost and speed against output quality; neither mode is intended for full-body action, multiple speakers, or camera-led scenes.

Motion Control is not a text-only generator: it requires both a character image and a motion video. The image supplies appearance and scene elements, the video supplies the action, and an optional prompt can clarify additions or motion intent; versions 2.6 and 3.0 expose the same core workflow with different model revisions.

Kling O1 is a unified multimodal video model for generating new shots and modifying existing footage. It can work from a prompt alone, animate first and last frames, incorporate image references, follow a video reference for style or camera movement, or treat that video as the base for an edit while preserving its timing and optional original sound.

Kling 3.0 Omni can create a new clip, animate start and end images, use visual references, or edit an existing video from one model entry. Its structured multi-shot prompting is useful when a short sequence needs explicit shot boundaries, while reference roles determine whether source media supplies content, style, camera language, or a base video to edit.