Seedance 2.5 Models

ByteDance Seedance 2.5 is the next generation of the Seedance AI video model suite, designed for unified multimodal video generation, editing, and extension workflows. It supports text-to-video, image-to-video, video editing, video extension, native 30-second generation, multilingual audio, and up to 50 multimodal references, including images, videos, and audio clips.

Built for cinematic content production, social media automation, advertising creatives, product videos, storytelling, and scalable AI video applications, Seedance 2.5 combines long-form generation, precise editing control, strong character consistency, native audio-video co-generation, and reference-based creative guidance in one production-ready model family.

Seedance 2.5 Series — Generation, Editing, and Extension

Text-to-Video Generation:

Generate cinematic videos up to 30 seconds from text prompts with multi-shot scene transitions, realistic motion, synchronized audio, and strong prompt alignment.

Image-to-Video Generation:

Animate a start image into a fluid video clip while preserving subject identity, visual style, composition, and scene continuity. Optional end-frame guidance can be used for more controlled motion direction.

Video Editing:

Edit existing videos with natural-language instructions while preserving the original duration, aspect ratio, motion structure, and scene continuity. Suitable for precise replacements, refinements, style changes, and production-ready video adjustments.

Video Extension:

Continue an existing video with a newly generated segment while maintaining motion continuity, visual style, character identity, and narrative flow.

Turbo Tiers:

Use faster and more cost-efficient variants for text-to-video, image-to-video, and video-edit workflows when speed, iteration, and high-volume production are priorities.

Talking Avatar:

Create talking-avatar videos from a portrait image and audio input with synchronized lip movement, natural facial expressions, and stable character appearance. Designed for AI presenters, virtual hosts, product explainers, training content, education, social media videos, and other digital human workflows.

Key Features

Native 30-Second Generation:

Create longer single-pass videos with coherent motion, multi-shot storyboarding, smooth transitions, and stronger narrative continuity.

Native Audio-Video Co-Generation:

Generate dialogue, ambient sound, sound effects, and adaptive music together with video, with multilingual audio support across 10+ languages.

50 Multimodal References:

Guide identity, motion, environment, style, and sound using up to 50 references in a single request, including images, videos, and audio clips.

Precision Video Editing:

Apply instruction-driven edits to existing footage while keeping unchanged areas, timing, layout, and visual continuity stable.

Character Consistency:

Preserve facial features, clothing, identity, and visual style across frames, shots, and extended video clips through reference-based guidance.

Cinematic Camera Control:

Control camera movement through natural-language prompts, including push-in, pan, orbit, tracking shots, scene transitions, and director-style motion.

Seedance 2.5 Models on WaveSpeedAI give creators and developers fast access to advanced AI video generation, editing, extension, and digital human APIs with scalable workflows, flexible pricing, and production-ready cinematic quality.