Kling 4.0 R2V

• 🏢 Kuaishou's reference-guided video model in the Kling family
• 🆕 Fourth-generation Kling R2V, listed with a 2026 release date
• 🎯 Builds clips from reference imagery to keep subjects consistent
• 🌐 Part of a 4.0 lineup spanning text, image and reference inputs
• 🔧 Succeeds the earlier V3 4K reference-to-video variant
• 💬 Catalog-positioned for human motion and editorial scenes
• 👁️ Takes visual references rather than prompt-only input
• 📚 Independent or official benchmark figures were unavailable at writing

Kuaishou

Kuaishou Technology is a Chinese publicly traded company founded in 2011 by Hua Su and Cheng Yixiao, headquartered in Beijing's Haidian District and listed on the Hong Kong Stock Exchange. Originally known for its massively popular short-video platform —…

Kling 4.0 R2V is the reference-to-video member of Kuaishou's Kling video family. Instead of starting from a prompt alone or a single source frame, a reference-to-video model takes reference imagery — characters, objects or settings — and generates motion around that material, so the same subject can recur across shots. The catalog lists its release date as 24 October 2026.

Within the 4.0 generation it sits alongside the text-driven Kling 4.0 and the frame-driven Kling 4.0 I2V image-to-video variant, which cover the other two common input modes. Its direct same-family predecessor is Kling V3 4K R2V, the third-generation reference-to-video entry; earlier reference-capable Kling releases include Kling O3 Pro R2V and Kling O3 Standard R2V.

Venice's catalog describes the Kling family broadly as cinematic in quality and well suited to human motion and editorial scenes, which maps naturally onto reference-guided work: advertising spots, short narrative beats and storyboard-to-clip workflows where keeping a face, costume or product recognisable across a sequence matters.

At the time of writing, no primary Kuaishou model card or independent top-evaluator results were available for this specific variant, so no performance figures or generation-over-generation quality claims are stated here. Users comparing it with the V3-era reference models should test on their own reference material.