AI Vibe Filmmaking Workflow with Storyboard Grids

> idea/creative direction: me, a human > 5×5 storyboard grids: Midjourney > ai video: FLUX3 at 720p > original soundtrack: Suno v6 > assembly and editing platform: DaVinci Resolve 21.1 > shot detection, beat matching, editing: GPT-6 Astra-light Build 5 Midjourney storyboard grids covering the four stages of the sequence: pure nostalgia, gradual contamination of everyday life, fear and grief, nightmare. Each grid should have 20–25 shots and a locked 1978 great american summer semantically dense visual language like "photochemical 35mm handheld footage" and "vintage cooke speed panchro spherical primes". FLUX3 doesn't allow reference images, so use each grid as visual guidance and aesthetic seeding. Prompt the model to show the grid for 0.5 seconds, then use it as a storyboard for the rest of the shots, treating each cell as a separate shot lasting 0.5–1.2 seconds, randomizing length selection. For Suno v6, use image and video references in simple mode. Describe a song style you have already experimented with. Generate tons of songs to get something directionally accurate, then iterate down more granularly to the final sound. Keep some phonetic structure from generated lyrics but rewrite every word to add human emotion matching the concept. For editing in DaVinci Resolve 21.1 with GPT-6 Astra-light: first split master videos into individual shots. Use Python + SciPy to detect beats and transitions in the song, convert them into 24 fps timeline markers in Resolve, then check everything with FFmpeg. Give Astra the editing style and vibe, with freedom to cut clips further to fit the music flow and video progression, slow them down, speed them up or do whatever worked. Give the LLM a clear goal and iterate on cuts.