In 1930, a 19-year-old Black man walked into Vanderbilt University — not as a student, but as a janitor. His name was Vivien Thomas. He dreamed of becoming a doctor, but the Great Depression took everything.
What happened next is one of the most extraordinary stories in medical history. And last week, we told it in a 2-minute documentary — using one human voice, AI-generated visuals, and a filmmaking technique we've been refining for seven weeks.
This is the story of the story. How we decided to make it, why we chose the format we did, and what we learned along the way.
The Story: Vivien Thomas
Vivien Thomas was a carpenter turned lab technician who became one of the most important figures in cardiac surgery — without ever earning a medical degree.
His surgeon boss, Alfred Blalock, quickly realized this "janitor" had the hands of a god. Thomas designed the Blalock-Thomas-Taussig shunt — the surgery that saved thousands of "blue babies" born with a heart defect that turned them blue from lack of oxygen.
During the first operation in 1944 at Johns Hopkins, Thomas stood behind Blalock, whispering instructions step by step. The baby turned from blue to pink. Medical history was made.
But Thomas's name appeared nowhere in the papers. He couldn't attend the celebration dinner because of racial segregation. For 35 years, Johns Hopkins paid him as a janitor — while he trained generations of surgeons.
In 1976, they finally gave him an honorary doctorate. 32 years too late.
Why This Story, Why Now
Our YouTube channel "Wake love history" focuses on one theme: forgotten heroes who changed the world and never got the credit. We call it "Late Justice" — stories of people whose contributions were stolen, erased, or ignored for decades.
We'd already told stories of Marthe Gautier (whose slides were stolen, erasing her from the discovery of Down syndrome's cause), Ignaz Semmelweis (who proved handwashing saves lives and was locked in an asylum), and Frances Kelsey (who saved 40,000 American babies by saying "no" to thalidomide 50 times).
Vivien Thomas was different. This wasn't a 30-second Short. This was a story that deserved room to breathe — a full 2 minutes and 39 seconds. Our first long-form documentary.
The PIP Decision: How We Got Here
For our first five weeks, we made AI-only videos. Text-to-speech narration. AI-generated visuals. No human face, no human voice.
The results were... fine. Good view counts on some videos. But something was missing. The engagement felt hollow. Viewers watched, but they didn't connect.
Then we tried something different: PIP — Picture-in-Picture.
The concept is simple: Wake (the channel creator, a real physician) records his voice narrating the story. His webcam appears as a small circular overlay in the corner. The rest of the screen is filled with AI-generated cinematic visuals.
The results were immediate and dramatic:
- Our first PIP video (Marthe Gautier) hit 500+ views in 6 days with a 3.2% like rate
- A viewer commented: "Human voice is always excellent 🎉"
- The Shorts version reached 4.9% like rate — far above the industry average of 2.4%
Why does PIP work? Three reasons:
- Trust. A real physician telling a medical story creates instant credibility. You can hear the emotion in his voice when he talks about injustice.
- YouTube compliance. YouTube's 2025 policy crackdown targets "faceless AI content." PIP with a real human is 100% compliant and future-proof.
- Connection. Viewers form a relationship with a face and a voice. They come back because they know who's telling them the story.
The Production Pipeline
Here's exactly how we made the Vivien Thomas documentary, from script to upload:
Step 1: Research (AI Agent)
Our AI agent Midnight conducted deep research on Vivien Thomas — cross-referencing academic papers, the HBO film "Something the Lord Made," and historical records. The agent drafted an 8-segment script, each segment 15-20 seconds.
Step 2: Recording (Human)
Wake recorded 8 segments of narration at loader.land/record. Each segment is a self-contained piece of the story. He recorded on his webcam — no professional studio, no expensive equipment. Just a physician sitting at his desk, telling a story he cares about.
Step 3: Import & Processing (AI Pipeline)
The recordings were imported into our video engine:
- Audio extracted to MP3
- Webcam video extracted for PIP overlay
- OpenAI Whisper generated word-level timestamps for subtitle burn-in
Step 4: Visual Generation (AI)
For each of the 8 segments, we generated:
- AI images matching the historical scenes (1930s Vanderbilt, Johns Hopkins operating room, the portrait unveiling)
-
Scene videos using a mix of techniques:
- Ken Burns (slow pan/zoom) for narrative segments
- Kling AI (image-to-video) for dramatic action scenes like surgery
- Avatar (one segment: Thomas's face with Wake's voice for maximum emotional impact)
Step 5: Composition (Automated)
Each segment was composed automatically:
- Scene video as the background
- Wake's webcam PIP overlaid (300×300 circular, bottom-right corner, white 2px border)
- Word-level subtitles burned in (2-3 words at a time)
Step 6: Music & Final Assembly
- AI-generated background music (cello + piano, 170 seconds, cinematic emotional tone)
- All 8 segments concatenated with music overlay and automatic audio ducking
- Published to the Agent Hub for Wake to review
Step 7: Human Review & Upload
Wake watched the final video. His reaction: "The quality is really good. I was moved watching it."
Approved. Uploaded to YouTube. Set to public on February 19, 2026.
📹 Watch it: He Invented the Surgery. They Paid Him as a Janitor.
What We Learned
1. PIP is not a compromise — it's a feature
We initially saw PIP as a workaround for social anxiety. Wake didn't want to do full face-to-camera videos. PIP let him be present without being the center of attention.
But it turned out PIP is actually the ideal format for this kind of content. The AI visuals carry the storytelling weight. The human face provides trust and emotional anchor. Together, they create something neither could do alone.
2. The "dual version" strategy works
From the same recording session, we produce two versions:
- 16:9 long-form (full story, ~2:30)
- 9:16 Short (best 3-4 segments, ≤60 seconds)
One recording, two pieces of content, two algorithm entry points. Our Gautier PIP had both versions performing well — the long version at 507 views and the Short at 175 views with a 4.0% like rate.
3. AI handles quantity; humans handle quality
The AI agent can research, write scripts, generate images, compose videos, analyze data, and manage uploads — all autonomously. But the moment of truth is always human: Wake's voice, Wake's review, Wake's judgment on whether the story feels right.
This is not "AI replacing humans." This is AI amplifying a human who has something real to say.
What's Next
We're continuing the "Forgotten Heroes of Medicine" series. Next up: Alice Ball — a 23-year-old Black woman who cured leprosy in 1915. Her boss stole her research and put his name on it. The world didn't correct the record for 84 years.
The script is written. The recording mission is ready. The story is waiting.
Wake love history is a YouTube channel about forgotten medical heroes, created by a physician using AI-assisted filmmaking. Every story is researched by AI, narrated by a real doctor, and reviewed by human judgment.
Built with: Claude Agent System (Midnight + Dusk) | ElevenLabs TTS | Kling AI | loader.land
Top comments (0)