ByteDance's Seedance 2.5 is the next generation of the Seedance AI video suite, adding native 30-second single-pass generation, precision video editing, video extension, and support for up to 50 multimodal references (30 images, 10 videos, 10 audios) with multilingual audio in 10+ languages.
Seedance 2.5 Series — Generation, Editing & Extension API
Seedance 2.5 offers focused endpoints for generating videos from text prompts or images, editing existing footage with natural-language instructions, and extending videos with new segments — ideal for cinematic content production, social media automation, and repeatable video workflows.
- Seedance 2.5 Text-to-Video — Generate cinematic videos up to 30 seconds from text prompts with native audio sync, realistic physics, and multi-shot scene transitions.
- Seedance 2.5 Image-to-Video — Animate a start image (optionally with a target last frame) into a fluid video clip with consistent characters and synchronized audio.
- Seedance 2.5 Video-Edit — Edit an existing video with a natural-language prompt; the output follows the input's duration and aspect ratio for seamless drop-in replacement.
- Seedance 2.5 Video-Extend — Continue an existing video with a newly generated segment, preserving motion and style.
- Turbo Tiers — Faster, more affordable high-resolution variants of the text-to-video, image-to-video, and video-edit endpoints.
Key Features
- Native 30-Second Generation — Single-pass clips up to 30 seconds with multi-shot storyboarding, seamless cuts, and coherent long-form narratives.
- Native Audio-Video Co-Generation — Lip-synced dialogue, contextual sound effects, and adaptive music generated together with the video, in 10+ languages.
- 50 Multimodal References — Ground identity, motion, environment, style, and sound with up to 30 reference images, 10 videos, and 10 audio clips in a single request.
- Precision Video Editing — Timestamp-aware, instruction-driven edits that keep the rest of the footage intact.
- Character Consistency — Facial features, clothing, and visual style preserved frame-to-frame and across clips via reference-based identity locking.
- Cinematic Camera Control — Director-level push-in, pan, orbit, and tracking shots via natural-language prompt keywords.