ByteDance Seedance 2.5: Video Creation Model
ByteDance's Seed team launched the new-generation video creation model Seedance 2.5 on 31 July 2026: building on the unified multimodal audio-video joint-generation architecture of Seedance 2.0, it centers on foundational and reference-based generation and makes breakthroughs in long-form storytelling, multimodal reference and editing; ByteDance says it can generate up to 30-second high-quality audio-video clips in a single pass with multi-round extensions, accept up to 30 images, 10 video clips and 10 audio clips as references in one pass, and offer timestamp-level targeted audio and video editing; at launch it was rolling out on Jimeng AI, Doubao Pro and other platforms, with API access coming via BytePlus ModelArk.
Tool Overview
Features, steps and FAQ below
Features
- ✓ 30 seconds per generation with multi-round extensions: ByteDance says Seedance 2.5 generates 30-second high-quality audio-video clips in a single pass and supports multiple rounds of extension, improving shot transitions and scene changes for stronger continuity in longer videos
- ✓ A complete story in one take: ByteDance says the model organizes multiple logically connected shots within 30 seconds so a story unfolds through setup, development, turning points and resolution rather than simply extending a single moment, letting users produce multi-minute content with a consistent audiovisual language
- ✓ Fully upgraded multimodal referencing: ByteDance says users can input up to 30 images, 10 video clips and 10 audio clips per pass, with strengthened reference capabilities including clay render, motion and creative references to grasp intent across multiple subjects, scenes and shot transitions
- ✓ More precise and stable editing: ByteDance says it offers timestamp-level control for targeted audio and video editing and enhances advanced features such as green screen, camera perspective and reference-based editing for professional fields like film and advertising
- ✓ More realistic texture and industry adoption: ByteDance says it systematically optimizes object textures, skin and eye features, lighting and color saturation and minimizes uncontrolled subtitles and background music, while exploring use in education, manufacturing, embodied intelligence and autonomous driving
How to Use
- In Jimeng (web) video generation, select Seedance 2.5
- In Doubao Pro video generation, select Seedance 2.5
- Provide multimodal references such as text, images, video or audio (up to 30 images, 10 videos and 10 audio clips per pass) and extend across multiple rounds as needed
- For API access, watch for availability via BytePlus ModelArk, which ByteDance says is coming soon
FAQ
What is Seedance 2.5?
ByteDance's Seed team says Seedance 2.5 is its new-generation video creation model that, building on Seedance 2.0's unified multimodal audio-video joint-generation architecture, centers on foundational and reference-based generation.
How long can its videos be?
ByteDance says Seedance 2.5 generates high-quality 30-second audio-video clips in a single pass with multi-round extensions, so users can produce multi-minute content with a consistent audiovisual language.
What reference materials does it support?
ByteDance says users can input up to 30 images, 10 video clips and 10 audio clips as reference materials in a single pass.
How capable is its editing?
ByteDance says it offers timestamp-level control for targeted audio and video editing and enhances advanced features such as green screen, camera perspective and reference-based editing.
Where can I use it?
ByteDance says Seedance 2.5 is rolling out on Jimeng AI, Doubao Pro and other platforms, with API access coming via BytePlus ModelArk.