Seedance 2.0 R2V

• 🏢 ByteDance Seed's flagship video family, reference-to-video endpoint, released March 2026.
• 🆕 Accepts combined text, image, video and audio references in one generation.
• 🎯 Prompts can point at specific references for composition, camera language, motion rhythm.
• 🔧 Adds video editing and extension: change clips, characters, actions, continue shots.
• 💬 Native joint audio-video generation with dual-channel audio, up to 15 seconds multi-shot.
• 📏 Upgraded instruction following and subject consistency versus earlier Seedance generations, per ByteDance.
• ⚡ Lighter Fast and Mini tiers, plus a newer 2.5 R2V, sit alongside it.
• 👁️ Directorial control over performance, lighting, shadow and camera movement.

ByteDance

ByteDance is a Chinese internet technology company headquartered in Beijing's Haidian district, with an associated variable-interest entity incorporated in the Cayman Islands. Best known as a consumer internet operator at enormous scale, it has also become a…

Seedance 2.0 R2V is the reference-to-video ("R2V") entry point into ByteDance Seed's Seedance 2.0 generation, released alongside the text-to-video and image-to-video variants Seedance 2.0 and Seedance 2.0. Rather than working from a prompt alone, it takes multimodal "all-round reference" input — text, images, videos and audio combined — and generates output that follows referenced elements such as visual composition, camera language, motion rhythm and sound characteristics; ByteDance notes it can even work from text-based storyboards.

The generational story is about control. ByteDance describes Seedance 2.0 as a substantial upgrade in controllability over its predecessors, with fully upgraded instruction-following and consistency, holding subjects stable across complex stories with rich character interaction and detailed action descriptions — a step beyond what the earlier Seedance 1.5 Pro line offered. The 2.0 release also introduced capabilities absent from prior versions: targeted editing of specified clips, characters, actions or storylines, and video extension that continues a shot from a prompt. Output is 15-second multi-shot audio-video with dual-channel audio, generated jointly rather than dubbed afterwards.

Within the 2.0 tier, this base R2V model sits above the cheaper, quicker Seedance 2.0 Fast R2V and Seedance 2.0 Mini R2V variants, and has since been followed by Seedance 2.5 R2V. It pairs naturally with ByteDance's image models such as Seedream V5 Pro, which can supply the reference stills that drive character and style consistency.