The video generator works in two modes. "Photo to video" brings an uploaded picture to life: a person smiles and turns their head, waves roll onto the shore, the camera glides around the scene. "Text to video" builds a clip from scratch based on a description — characters, setting, camera movement and mood. Most models generate sound right along with the picture: speech, music and ambient noise.
All of 2026's leading video models are available in one interface. MiniMax H3 Turbo is the fastest and most affordable, with a clip ready in about a minute. MiniMax H3 is especially good at animating portraits. Kling 3 and Kling 3 Pro deliver cinematic camera movement and lighting. Google's Veo 3.1 offers realistic physics. Gemini Omni Flash is Google's new model with precise prompt adherence and speech. Alibaba's Wan 3.0 excels at building narrative scenes with shot changes. Grok Imagine 1.5 from xAI and Kling 4.0 can produce clips up to 15 seconds in a single generation. Seedance 2.5 from ByteDance is a premium model for complex action and speech with lip sync.
Duration is 5 or 10 seconds (Gemini Omni Flash also offers 8, Veo 3.1 offers 4, 6 or 8 seconds, and Grok Imagine and Kling 4.0 go up to 15), resolution is 720p or 1080p, with 16:9, 9:16 and 1:1 formats. Generation happens entirely in the cloud: no graphics card or software needed. The finished MP4 can be extended, edited by description, upscaled to 4K, or published right away.