Seedance 2.5 on Luma doubles the continuous video generation limit to 30 seconds and removes the need for stitching, closing a key gap for solo video creators.
What changed
Luma AI has released Seedance 2.5, now live on the Luma platform. The update allows users to generate a single 30-second video clip in one pass, with no seams or resets between segments. This is roughly double the typical 15-second limit of previous models, which required splicing and often led to visible artifacts or character drift.
Evidence and capabilities
- 30-second continuous video generation in a single run
- Support for up to 50 reference inputs (images, video, audio, style), up from 12
- Native audio generation, built and synced with the video
- Live style swap (e.g., live-action, illustration, anime) with consistent composition
- Targeted fixes: adjust or add objects without re-rendering the whole scene
- Improved prompt adherence and native support for 10+ languages
Workflow impact
This release reduces manual editing and post-processing for longer clips, and the expanded reference and style controls enable more complex and tailored outputs. Indie builders can now experiment with richer, longer-form content without technical workarounds.
Practical guidance
Test Seedance 2.5 for your use case, especially if your workflow previously required stitching or suffered from scene drift. Evaluate whether the new reference and style features improve creative control. Pricing and API access remain unchanged as of this release.
Limitations
All performance and quality claims are from Luma AI. Independent benchmarks and user feedback are needed to confirm real-world reliability and output quality.