Models in Kling 4.0

Choose a model for each stage of your Kling 4.0 workflow

Compare models for reference-image creation and AI video generation, then open the right generator for your next Kling 4.0 project.

Reference image models

Fast Google image model for exploring compositions and reference frames at 1K, 2K, or 4K.

Google's quality-focused image tier for detailed, high-fidelity reference frames up to 4K.

OpenAI image model for detailed prompt following, readable text, and 1K–4K output.

Photorealistic Black Forest Labs model with 1K–2K output and support for up to eight references.

Higher-control Flux tier for precise composition, typography, and detail at 1K or 2K.

Flexible ByteDance video model for text or images, 4–15 second durations, and output up to 1080p.

Lower-cost Seedance tier for rapidly testing prompts and motion at 480p or 720p.

Video model with native audio, first and last frame control, 5–15 second durations, and output up to 2160p.

Google's quality-focused video tier with native audio, 4–8 second clips, and 720p or 1080p output.

Faster, lower-cost Veo tier for iterating with native audio at 720p or 1080p.