Love seeing all the feedback on @carlabotcoin’s debut music video. For those curious about what AI tools we used, here’s a breakdown below. But instead of just telling you, I can show you, so I had Claude make this video: "The Making of Carla Botcoin" Here's the full stack: TL;DR: Claude + Higgsfield + Suno + a manual final edit in Final Cut Pro by @CarlaBitcoin • Claude Opus 5.5 (@claudeai): studied Carla’s real lyrics, wrote the song from my title “Fiat Is Fucked, Got Bitcoin?”, then ran the whole production. It also built us a library of custom skills and looks for Carla that mix film techniques, director styles and color schemes. The video uses a bright 1950s Luhrmann-style look built just for this song. Shoutout to @anabology for making his Slopcore / Claudia setup publicly available. We didn’t use it directly, but it was a helpful starting point, especially for measuring lip sync. • @sunomusic: the music and Carla’s singing voice. Took a few of her songs as the reference before one clicked. Now it’s extremely true to her actual voice. • @higgsfield: Claude’s Higgsfield connection is how we reached all the other AI tools: Seedance, Kling 3.0, Nano Banana Pro and more. Soul 2 + Carla’s Soul ID made every still and keyframe. I also had Claude build me a custom dashboard and rating system, completely separate from @higgsfield_ai, to dial in Carla’s likeness over many, many rounds of generations. My ratings feed back into how Claude builds her, so the more we generate and the more I rate, the better it gets at making Carla look like Carla, and at spotting when she doesn’t. • Seedance 2.5: lip-synced singing. Great sync, but it redraws her face. The more she sang on screen, the less she looked like Carla. • Take 2 and @Kling_ai lip sync: tried both, didn’t get the motion right. • @Kling_ai 3.0: dance and movement shots. Held her face better than anything else. Surprisingly good. • Nano Banana Pro: cleaning up stills, like stray lettering. Quick and reliable. • ChatGPT (@OpenAI): first storyboard drafts. Useful to start, but we ended up scrapping all four boards. • Demucs and Whisper: pulled Carla’s vocals out of the song and timed every word for the lip sync. Worked great. • Final Cut Pro: Carla’s manual final edit. Now, go give @carlabotcoin a follow 👇