Vidu Reference

Q3 Mix is the long-form, high-definition tier; Q1 is the fixed-length tier with motion control. Q3 Mix makes 1 to 16 second clips up to 1080p with synchronized native audio, with pricing that scales by duration and resolution. Q1 has a fixed clip length and lower maximum resolution, no native sound effects or dialogue, but offers a Movement Amplitude setting and optional background music at a fixed cost.

Use Q3 Mix. It is built for multi-entity storytelling, supports up to four reference images, runs up to 16 seconds at 1080p, and generates footsteps, ambience and dialogue in the same pass. Expect longer render times on 1080p long clips.

Pick it for lightweight reference tests or legacy workflows that don't need feature fusion or native audio. It is the baseline engine, so interaction physics and prompt adherence are less capable than Q3 Mix.

Use one to four references. Clean, well-lit portraits or object turnarounds on neutral backgrounds give the sharpest likeness. Avoid low-resolution or heavily cropped images, and avoid mixing conflicting lighting and art styles unless you want a deliberate fusion.

It struggles with fine hand work like playing piano or tying knots, legible in-frame text and typography, and abrupt scene cuts inside one take. Fast large-amplitude camera moves can soften background texture, and three or more closely overlapping subjects can occasionally blend.

Yes, with Q3 Mix, which generates synchronized foley, ambience and vocalizations natively when Generate Audio is on. Describe sounds or dialogue in your prompt to steer it. Q1 does not make sound effects or dialogue, but can add background music; it was created by ShengShu Technology.