AI Music Video Generation Workflow

Turn mp3 into music video using invideo agent two WORKFLOW: Step 1 - I uploaded the mp3 song, one reference image per character, and optionally lyrics. Step 2 - Agent two analyzes the song It maps BPM, verse/chorus timestamps, and vocal entrances down to word-level lyric timing - and makes the audio the master timeline. Step 3 - then it builds the world. It generates a location sheet (the world bible: environment, props, lighting) and locks the characters to your reference images so faces never drift. Step 4 - Agent Two slices the song. It trims the track into exact 15-second segments - 0–15, 15–30, 30-45… each one the timing source for its own video block. Step 5 - it generates chained sequences For every block it pulls the previous clip's final frame as continuity, pairs it with the character refs and the audio slice, writes the director-level prompt, and generates. Lip-sync comes out matched to the real vocals automatically. Step 6 - it automatically chooses Hailuo minimax H3 model for music video generation which is good at character music lipsync matching the rhythm. Step 6 - if something went wrong like, wrong seat, lighting drift? Agent two rewrites the prompt, locks what drifted, and regenerates only that block. Step 7 - once all clips generated, It assembles all clips on its editor called slate, mutes their baked-in audio, lays the original MP3 as one continuous bed, adds the fades, and renders.