Seedance 2.0 & Midjourney 8.2: Advanced AI Video & Image Generation

Seedance 2.0 & Midjourney 8.2: Unleashing Advanced AI Creation

This update brings significant enhancements to ByteDance's Seedance 2.0 family (Fast/Pro/Mini), introducing reference video and audio capabilities, along with more flexible duration controls. Additionally, Midjourney 8.2 is now available in our model library. Whether you're creating short videos, product animations, or a series of visually consistent posters, these updates are worth exploring.

Seedance 2.0: Truly Referencing Your Input

Previously limited to reference images, Seedance 2.0 now supports video and audio inputs:

• Reference Images: Up to 9 images.

• Reference Videos: Up to 3 clips (total length ≤ 15 seconds).

• Reference Audio: Up to 3 clips (each 2–15 seconds, requires at least one reference image or video).

• Simultaneous Use of Start/End Frames and Reference Materials: You can now lock the opening and closing frames while using reference images to maintain subject consistency.

• Flexible Generation Duration: Choose any duration between 4–15 seconds, moving beyond fixed options.

• "Generate Audio" Toggle: Disable audio generation for silent videos when sound is not needed.

Here are two immediate use cases:

1. Use a favorite camera movement or character motion video as a reference. Seedance 2.0 will apply the motion and pacing to your own characters or products, with AI reinterpreting the visuals based on your input.

2. Specify opening and closing frames, then add product reference images. This locks the composition for the intro and outro while ensuring product details remain consistent throughout the generation.

For advanced control over camera language, refer to Seedance 2.0 Text-to-Video Prompt Tutorial.

Pricing is now transparent: Videos with reference clips are billed based on the total input + output duration. The system automatically reads the actual duration after upload, providing an estimate like "8 seconds input + 7 seconds output" before generation.

TierResolutionWithout Reference VideoWith Reference Video (Input + Output Total Seconds)Fast480p / 720p20 / 42 points/sec12 / 25 points/secPro480p / 720p / 1080p24 / 52 / 128 points/sec15 / 32 / 78 points/secMini480p / 720p12 / 26 points/secSame price × Total seconds

Midjourney 8.2: Classic MJ Aesthetics, 4 Images at Once

The new flagship model, Midjourney 8.2, is now live. It retains the familiar MJ aesthetic, generating 4 images per prompt at 1K (23 points) or 2K (35 points). It supports image prompts, style references --sref, and native parameters like --stylize, --chaos, --raw. Prompts must be in English (with built-in AI translation assistance).

A practical application: Use the same --sref style code to batch-produce a set of posters or cover images with a unified tone. Change the subject and composition while the style code ensures consistency. For parameter selection guidance, refer to Midjourney 8.2 Prompt Guide.

Infinite Canvas Integration: Video-to-Video Generation

The Infinite Canvas (Beta) also supports these updates. Select a single video node to initiate "video-to-video" generation, using a pure video as a reference for re-generation. Cost estimates now include a breakdown of actual durations, so you know the point cost before generation is complete.

Try It Now

Whether you want a reference video to drive new actions or to lock down a visual style with --sref, these capabilities are now available in Genvizu. Open Genvizu AI Infinite Canvas.