What is the maximum single-generation duration?

ByteDance Seed documents up to 30 seconds per generation, with multiple rounds of extension.

Which references can it accept?

The official release lists up to 30 images, 10 video clips, and 10 audio clips in one generation.

Is Seedance 2.5 available in Visnova?

This article does not announce an integration. Check the current product interface for models available to your account.

Seedance 2.5: documented features and a practical review workflow

ByteDance Seed announced Seedance 2.5 on July 31, 2026. Its official release describes single-pass videos up to 30 seconds, multiple rounds of extension, up to 30 image references, 10 video references, and 10 audio references per generation, plus timestamp-level editing. See the official release and model page.

1. Capabilities documented by ByteDance SeedVISNOVA AI

Longer scenes: one generation can produce up to 30 seconds, and clips can be extended in multiple rounds.

Multimodal references: a request can include up to 30 images, 10 video clips, and 10 audio clips. The model can use them to guide subjects, motion, camera work, and sound.

Targeted editing: timestamp controls can direct edits to portions of audio or video. The release also describes green-screen, camera-perspective, and reference-based editing.

These are model-level capabilities. Actual limits and controls depend on the service or API offering the model.

2. A practical creation workflowVISNOVA AI

1Check availability and settings. Choose a service that lists Seedance 2.5 and confirm its current duration, reference, resolution, and export options. ByteDance's release named Jimeng AI and Doubao Pro for rollout and said API access would follow. This article does not announce a Visnova integration; check the current Visnova interface for available models.

2Prepare references with permission. Use clear product and scene images, concise motion or camera references, and audio that you are authorized to use. State which details should remain stable and what should change.

3Generate, inspect, and revise. Review the whole clip for product shape and color, logos and text, hands and movement, continuity, captions, and audio. Use targeted edits where available, then inspect the edited clip again.

3. Review before commercial useVISNOVA AI

Generated video can alter product details, text, people, or sound. Compare the result with the real product and approved brand assets. Proofread captions and claims, confirm rights to references and audio, and check the rules of the destination platform. Do not treat a realistic-looking output as proof of accuracy or permission.

Seedance 2.5: documented features and a practical review workflow

1. Capabilities documented by ByteDance SeedVISNOVA AI

2. A practical creation workflowVISNOVA AI

3. Review before commercial useVISNOVA AI

Explore related topics

FLUX Video Upscale: Features, Output Limits, and Review Tips

GPT-6 Astra for Fashion Workflows: Capabilities and Limits

Qwen-Image-3.0 for E-Commerce Visuals: Features and Review Tips

Frequently asked questions

Create a fashion visual draft with AI

Scan to Contact Fashion Advisor

Join AI Fashion Creator Group