Faceless TikTok with AI voice: 90 days, 340 videos, $2,180 — full breakdown

Running this as an experiment in how far a fully automated pipeline gets before quality becomes the bottleneck. Answer: about six weeks.

The pipeline
Script generated from a trending-topic feed, voice synthesised, stock footage matched by keyword, captions burned in, posted by API. Human touches the script for about 40 seconds per video.

Output and result
340 videos in 90 days. 11.4M total views. 84k followers. Revenue: $1,240 creator rewards, $940 affiliate.

Where it broke

• Weeks 1-6: views climbing, everything fine
• Week 7: average views dropped 70% with no change to the pipeline. Comments started saying the videos felt generated
• Weeks 8-12: recovered to about half the peak only after I started rewriting the hook by hand on every video

What I think happened
Not a penalty. The audience got bored of the format faster than the algorithm did. The videos that survived the drop all had a hook a human wrote.

Would I do it again
Yes, but with the human at the front rather than nowhere. $2,180 against roughly $290 of API and stock costs is a fine return. $2,180 for 340 videos is $6.40 a video, which is a terrible wage if you count my time honestly.

Week 7 being the wall is interesting — that is roughly when a regular viewer has seen you 20 times. Format fatigue, not platform punishment.

The $6.40 a video line is the most honest thing in this section this month. Most automation posts stop before the division.

It was tempting to stop before it, I will admit.

Did you try the same pipeline across three accounts in different niches? Would tell you whether it is audience fatigue or your particular audience.

Running that now, two accounts, 30 days in. Early read is the same curve, different height, which points at the format.

Worth noting the comments told you before the metrics did. Reading your own comment section is underrated analytics.

Disagree with half of this but the half I agree with is worth the read.