"Almost Famous" Podcast

What it is

Almost Famous: Lost Heroes & Forgotten Legends is a documentary podcast about brilliant people whose contributions were overlooked, suppressed, or quietly lost to history. Rosalind Franklin's work on DNA. Klara von Neumann's pioneering computer programming. Lucy Parsons. Bass Reeves. The enslaved inventors of the cotton gin. Harriet Burns, Disney's first female Imagineer. The show focuses on the figures who almost changed the world, and the forces (gender bias, class oppression, political agendas, industrial espionage) that systematically erased them.

Why I made it

I run a small media-production lab. I hate the sound of my own recorded voice. There's no budget for voice talent or freelance editors. A traditional production setup for a long-form documentary podcast was out of reach.

That's the brief in the form it actually came in. Not "let's experiment with AI" but "here's something I want to make, and the only honest path to making it runs through tools that didn't exist a few years ago."

So Almost Famous became, by necessity, a working test of a particular division of labor: human judgment at the checkpoints, AI everywhere else.

The pipeline

Nine steps, three human checkpoints, the rest automated. Each episode is tracked in a Google Sheets workbook from concept to publish.

1. Concept generation via Perplexity AI.
2. Concept selection. HITL (Human in the Loop). The editorial mission lives here. The whole show fails if the wrong figures get picked, so I review every candidate by hand.
3. Knowledge corpus building via NotebookLM and Perplexity.
4. Script development via NotebookLM and Gemini.
5. Fact-check and sensitivity review. HITL (Human in the Loop). Perplexity flags claims and sources. I read every script for accuracy, sensitivity, and bias before audio generation runs. Minimum three academic sources per claim.
6. Audio generation via ElevenLabs TTS, with me assembling the final mix in DaVinci Resolve.
7. Show notes generation.
8. Editorial review. HITL (Human in the Loop). Final pass before publish.
9. Publishing to Acast, which fans out to Spotify, Apple Podcasts, YouTube, iHeart Radio, Deezer, and Amazon Music.

Three human checkpoints. The rest is plumbing.

What this taught me

This pattern (human judgment at the checkpoints, AI everywhere else) isn't only a strategy for established creators protecting a premium. It's also an equity story. Almost Famous wouldn't exist without TTS. A single person with no budget can now produce a long-form documentary podcast series that used to require a production team. AI didn't replace a human team here. It made production possible where nothing existed before.

The harder skill is knowing when AI makes sense and when it doesn't. Use AI ruthlessly where it earns its keep. Stop using it the moment human judgment matters more than throughput. Concept selection, fact-check, editorial review: those don't get automated. Everything else does. The challenge is the judgment call about which is which, and that judgment is the practitioner's job, not the model's.