
How I turned a 6-year-old film into 9.9M+ views in a week with AI
Facebook offered a $5,000 bonus for a five-part series. I took 2020 Nepal footage of a stray dog named Ghost and used Claude, ElevenLabs and Final Cut to ship five chapters and one 2:45 cut in a week. Here is the exact workflow, the numbers across every platform, and what broke.
Watch on every platform (Nov 19)
The source footage is my 2020 YouTube film, Hiking Alone in Nepal.
The numbers
The five-part series was the reason I started. The single 2:45 cut, edited on November 13 for Instagram, YouTube Shorts and TikTok, is what actually took off.

As of November 19:
- Instagram: 5.5M views (2:45 single cut)
- YouTube Shorts: 2,146,076 views (2:45 single cut)
- TikTok: 1.8M views (2:45 single cut, disqualified from Creator Rewards, $0 paid)
- Facebook: 530,445 qualified views across five chapters (total plain views on Facebook are over 900,000; only qualified views count toward the challenge)
Together that is 9.98M+ views in ten days from footage sitting on a hard drive since February 2020.
The source material
Most creators with a few years of uploads are sitting on stories that already proved they work. In February 2020 I hiked the Mardi Himal Trek in Nepal. On the second morning a small white stray dog I called Ghost started following me through the forest, through a whiteout, and up to High Camp, where he slept on my jacket.

I already knew the story resonated. When I cut a rough version for TikTok in January 2023, part two alone reached 6.2 million views. And the story had a real ending: after my hike, people online tracked down a photo of a woman at the Pokhara airport with Ghost on a leash, taking him home to the US.

When Facebook's $5,000 series challenge appeared in early November, I needed one arc that could carry five one-minute chapters without shooting anything new. Ghost was the obvious test.
How it was made
I am open about the AI pieces because the pipeline is the point of this experiment. It is my trek, my footage and my story; the tools let me ship five episodes and a full remix in a week while working on other projects.

- Scripts from the footage. Claude reviewed my 2020 film and its companion guide video, checked every fact against the timeline, and drafted five chapters of about 190 words each: the dog at breakfast, the forest, the whiteout, the top, and the goodbye. Every chapter opens with the same hook ("This dog followed me for days while hiking in the Himalayas") over a different first shot of Ghost.
- Voice clone. I generated the narration with my ElevenLabs voice clone from the five scripts. I listened through all five takes and kept them as generated rather than re-recording in a booth.
- The Final Cut feedback loop. This was the most useful part of the workflow. Claude assembled each chapter as a 9:16 vertical rough cut from my original footage and exported a Final Cut Pro XML file. I finished the first three chapters in Final Cut myself. Then Claude compared my finished timeline against its draft to measure what I changed. On Chapter 1 I moved 8 of 23 cut points and reframed about a dozen shots. By Chapter 2 that dropped to 4 cut points.
- Original score in code. Rather than digging through stock libraries, Claude wrote a simple score in code, note by note, with nothing sampled, timed so the shifts land on the story turns.
- The 2:45 remix. Once the five Facebook chapters were done, I stitched the whole arc into a single video under three minutes for Instagram, YouTube Shorts and TikTok.
The Facebook challenge
Facebook's bonus rules are straightforward: publish at least five episodes telling one story in order between November 3 and December 1, each over a minute long, and get 10,000 qualified views on every episode to unlock $5,000.
The five chapters went up between November 9 and 13. As of November 19, four of five have passed 10,000 qualified views. Only Chapter 1 is still short, with 2,835 qualified views left and 12 days to go.
The five Facebook chapters (qualified views, Nov 19)
The distribution was lopsided: Chapter 3 took off on its own (over 406,000 qualified views), while Chapters 1 and 2 lagged. To help the early chapters along, I shared them in a few Facebook groups about Nepal and dogs, pinned comments on the later chapters pointing viewers back to Chapter 1, and put about $100 in ads behind the first two.
I will update this page after December 1 with whether Chapter 1 cleared 10,000 and whether the $5,000 bonus paid out.
The TikTok twist
On TikTok, the 2:45 single cut passed 1.8 million views and made $0.
TikTok's Creator Rewards program marked it "Disqualified" with no explanation beyond a generic content policy notice. My best guess is that the system flagged either footage that already exists on YouTube or the AI-generated voice. I appealed, and I will note the outcome here if it changes. Either way, it is a good reminder not to build a workflow around a single platform's payout rules.
Takeaways for creators
- Your archive is underused. Six-year-old footage that already earned views on one format can run again if you repackage it cleanly for vertical video.
- One complete cut beats a five-part series for reach. The five chapters existed for Facebook's bonus rules. The single 2:45 video drove 95% of the views because people will share one complete story, not a cliffhanger.
- Diffing your edits teaches the model your taste. Handing an AI rough cut to Final Cut and then feeding your finished timeline back to compare cut points is what made the later chapters fast to finish.
- Series need an explicit way back to Part 1. Short-form feeds drop viewers into Chapter 3 or 5 at random. Pin a link to Chapter 1 on every later episode the day it goes live.
- Keep the pipeline cheap enough to survive $0 payouts. Because the whole production took a week alongside other work, TikTok's disqualification was an annoyance instead of a loss.








