Complete Theory: How to Make High-End AI Music Videos

1. Music Generation (Suno v6)

Beat-First Approach: Instead of generating the entire song at once, start by creating a purely instrumental beat to get the vibe right.
Remixing: Download that instrumental, upload it back into Suno as an audio reference, add your lyrics, and remix the style (e.g., adding specific tags like "K-pop girl group trap funk fusion").

Conceptualization: Give your characters a backstory, physical appearance, and personality before generating.
Face & Outfit Generation: Use an LLM (like ChatGPT/Claude) to structure your prompts. Generate the face first. Then, find outfit references (e.g., from Pinterest) and ask the LLM to write a prompt recreating that outfit.
The Ultimate Character Sheet: Combine the close-up image of the face and the full-body outfit image inside Hicksfield, using a structured prompt to generate a cohesive character sheet showing multiple angles.

Video Set Generation: For minimal sets, you can directly prompt Seedance 2.5 (e.g., "bright minimal white studio").
Image Set Generation (Recommended): For high-budget, complex sets, find reference photos, have an LLM reverse-engineer the prompt, and generate static set images using image models (like Cream 5.0 or Nano Pro) to use as your backdrops.

Dynamic Angles: Avoid static shots. Use distinct camera angles like Dutch angles, low angles, overheads, and ground-level shots.
Movement: Pair angles with movements like crash zoom, handheld, tracking shots, or orbits. (You can also add camera shake in post-production).
The "Hardcut" Technique: To get highly dynamic footage, use a prompt trick: command the AI to "cut to a completely different angle every second." This generates 15 completely unique angles in a single 15-second generation.
Genjutsu (Character Swap): If you get a perfect video sequence with a great camera movement but want a different character, use Hicksfield's Genjutsu feature to swap the subject while perfectly preserving the camera motion.

Stem Separation: Never upload the full track for lip-syncing. The AI will mistakenly try to make the character's mouth move to the beat, instruments, or background ad-libs.
The MP4 Trick: Extract only the Lead Vocal stems in Suno. Turn that isolated vocal track into an MP4 video file with a plain black screen, and upload that into Hicksfield as your audio reference. This forces the AI to lip-sync only to the main lyrics perfectly.

For quick social media promos, upload your Character Sheet, your vocal MP4 (up to 30 seconds), and a highly detailed prompt to Seedance 2.5 to generate a continuous 30-second music video clip all at once.