AI video tools keep moving at warp speed, and so does our “Directing the Future: Ethical AI Video for Filmmakers” course on MZed. In the newest module, which has just gone live, cinematographer and photographer Drew Geraci returns to Higgsfield AI, the all-in-one generative AI platform. This time, Drew puts one of the latest models, Seedance 2.5, to the test. What can you learn from his experience? Let’s find out!
In one of the previous course updates, Drew took a close look at Higgsfield AI’s interface as a whole. In the latest lesson, however, he narrows in on one specific model inside it: Seedance 2.5, built by ByteDance. According to Drew, he has been testing it for the past four to five months, and by his own account, it’s the most realistic AI video generator he’s used so far.
Seedance 2.5 is ByteDance’s current flagship video generation model. Integrated directly into Higgsfield’s interface, it can produce clips up to 30 seconds long in 4K, in aspect ratios ranging from vertical 9:16 to widescreen 21:9. Two features stand out in Drew’s demonstration. First, audio: rather than adding sound afterward, the model generates ambient noise, SFX, and music in the same pass as the picture, matched to what’s happening on screen. Second, reference-based consistency: you can feed it up to 50 images or clips of a person, product, or location, and it will keep that subject looking consistent across the entire sequence.
SOURCE:
