Lyrical Legend

Proof that AI Artist can write impressively themselves and only need AI to carry the parts that come after. A flex piece, sure, but also a reminder that hip hop is more than politics and heavy topics.

Comment

Inspiration

Song: A flex-piece response to a challenge but I also wanted to show pure craft without abandoning honesty. A temporary pivot from my usual political and social themes to celebrate technique, speed, and internal rhyme density. Visual Identity: A picture of a woman on a velvet chair on a mediterranean rooftop. The chair almost acts like a throne. The Cats carry a lighthearted backdrop with the piano play B-roll.
This Video was built specifically for the Chroma Awards with the help of several subscriptions provided be the sponsors during submission time

What it does

Delivers a high-velocity lyric performance that foregrounds technical writing, flow switches, and conversational punchlines. Presents my cohesive visual identity of Aidan Yagu: Functions as a calling card for my creative process where tools assist and the artist leads.

How we built it

Music: Lyrics made by hand, AI music gen through Suno shaped for cadence and emphasis. Look development and storyboards: Flux1 Kontext for key frames and art direction. Video generation:

This is where the patchwork comes in LTX Studio for broad sequences, their older model gave me ample ways to explore variety in visuals during the sponsor month. Kling via OpenArt for very strong "first-frame" adherence and framing control. Veo3 and Nano Banana for the true-to-prompt shots, often conditioned with Flux frames. Sora for the unexpected angles and variations beyond strict first-frame continuity. Sora excels in being a wildcard. Edit and grade: DaVinci Resolve Studio. Unified cine-neutral base, restrained grain, subtle vignette. Cuts land on bar onsets and consonant hits.

Challenges we ran into

Mask fidelity: the white head and centered line are hard for most models to keep consistent. Required strict anchors and negative prompts. LTX Support also confirmed masks in general to be a weakpoint. I had to find ways around, but I'm glad I did. Cross-model cohesion: despite the different models used it needed to feel cohesive. and lastly the performance feel: achieving energy without literal lipsync by cutting on rhythmic language cues

Accomplishments that we're proud of

A cohesive, award-ready music video assembled from diverse AI sources that still reads as one vision. A clear demonstration of technical lyricism that stays musical and conversational. A reusable visual bible for Aidan Yagu that travels across tools and projects.

What we learned

Use models for what they are best at: Kling for first frames, Veo3 and Nano Banana for fidelity and prompts adherence, Sora for discovery is still godsend, Flux for art direction, especially when you don't have access to midjourney. Lock audio first. Picture moves faster when the rhythm bed is final.

What's next for Lyrical Legend

maybe some video shorts and hopefully some viral success. but again, the project is done.

Built With

  • davinci
  • kling
  • ltx-studio
  • nano-banana
  • sora
  • veo3
Share this project:

Updates