H3 Timeline Pixaroma

Tell one long MiniMax H3 video as scenes: each scene has its own length, a few words for what happens, and, if it needs one, its own picture. Scene 1's picture is the start picture. A later picture is either a New look (the same shot with a change, like a hat or another dress; it fades in, so for a natural change write the action in the scene before, with no picture) or Add person (any photo of someone who joins; they can walk in). It replaces MiniMax H3 Reference to Video and H3 Long Video: wire model, clip and both VAEs in, and the model, positive and latent out to H3 Audio Sync and the sampler. Sing and Talk put your words into the tested EP41 prompts and use the audio input as the song or the voice (lock it with H3 Audio Sync); Act needs no audio and H3 makes the sound itself. The video is made in one run, in windows that follow your scenes, so each scene change blends over a couple of seconds with no cut or flash. A scene is 4 seconds or more. A person who joins stays for the scene they join in: write their arrival and what they do next in one scene, and keep that scene at about 13 seconds or less (a longer one is made in two parts and the person can slide between them). Works only with MiniMax H3.