$100 AI Music Video: Claude Fable 5 vs. GPT-5.6 Sol
Points and comments are a snapshot, not live.
An agentic harness gave Claude Fable 5 and GPT-5.6 Sol $25-$100 to autonomously produce music videos.
Two frontier models were given a song, budget, and tools to research, generate, edit, and assemble a full music video. Claude Fable 5 and GPT-5.6 Sol each ran at $25 and $100 budgets. GPT-5.6 Sol at $25 used an image-to-video pipeline and was the most inventive editor. Claude Fable 5 cost more in LLM tokens ($16.99-$25.05) but produced slightly more coherent output. None of the videos were great. Both models struggled with character and story consistency, took lyrics too literally, and rarely iterated on their edits. Neither touched Replicate, and $100 was likely too much budget.
What commenters are saying
Most commenters found the output videos poor, with several noting the absurd literalism of AI interpreting lyrics. Some defended literal music videos as a valid style, citing Weird Al or Vance Joy as examples. A few argued that better scaffolding (e.g., focusing on animation or cartoon styles) would yield better results. Others suggested the chosen video models were not state-of-the-art. One commenter claimed AI enables experts to produce work without external capital, framing it as a positive shift.