Seedance 2.5

• 🏢 ByteDance's flagship Seedance video family, reference-to-video route on Venice.
• 📏 Generates up to 30 seconds of audio-video in one pass.
• 🆕 Supports multiple rounds of extension for longer storytelling.
• 🎯 Reference control spanning clay render, motion, and creative references.
• 🔧 Timestamp-level editing for targeted audio and video changes.
• 👁️ Optimized textures, skin, eyes, lighting, and color saturation.
• 💬 Joint audio-video generation, not silent clips plus dubbing.
• 🌐 Sibling to ByteDance's Seedream image and editing models.

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 2.5 is ByteDance's current flagship video generation model, released in 2026, and this catalog entry exposes its reference-to-video route: you supply reference material alongside a prompt and the model composes a single coherent clip. ByteDance describes it as a next-generation audio-video joint generation model built for 30-second storytelling with precise reference control and editing capabilities. Audio is generated together with the picture rather than added afterward.

The clearest generational change over earlier Seedance releases is duration and continuity. ByteDance states that Seedance 2.5 produces high-quality 30-second audio-video clips in a single pass and supports multiple rounds of extension, with improved shot transitions and scene changes for stronger continuity in longer videos, plus what it calls notable gains in image, audio, and motion quality. Reference handling is also broader, with specific reference modes including clay render, motion, and creative referencing — clay render, for example, lets you define spatial structure, character poses, motion paths, and camera angles with textureless 3D models.

ByteDance also reports systematic work on the "artificial" look of AI video — object textures, skin and eye detail, lighting, color saturation — and fewer uncontrolled subtitles or background music, plus timestamp-level control for targeted edits.

Within Venice's ByteDance lineup, Seedance 2.5 can be used alongside the still-image and editing models: Seedream V5 Pro for still-image generation, Seedream V5 Lite for lighter generation, and Seedream V5 Pro Edit for preparing reference frames before animating them here.