Video Prompt Pixaroma

Writes a MiniMax H3 video prompt for you, on your own machine, using a small language model you already have. It replaces three separate workflows and about ten nodes with one. Type your idea in plain words, pick how long the video should be, and press Run. The node hands back a finished H3 prompt with all the fields and rules that model expects, plus the frame count to render it at. What it writes depends on what you wire in, and it switches by itself. Nothing connected means text to video. A first frame connected means it looks at that picture and animates it. Both a first and a last frame means it writes the journey from one to the other, and it joins the two pictures for you so they can never end up the wrong way round. The wording it follows lives in the settings, one for each of the three cases, and you can edit any of them and put the original back. The length choices live there too, because how much to write is the setting that changes the result most. Wire the frames output into your H3 node so the video is rendered at the same length the prompt was written for. Getting those two out of step is the easiest way to spoil a clip. Needs a vision model in your text_encoders folder, because the first-frame modes have to see the picture. Find it by searching for h3, minimax, prompt, llm, or write prompt.