The $0 TTS setup behind 60+ of my YouTube shorts
Every faceless-channel guide gives you the same shopping list, and the first paid item is always a TTS subscription at $5 to $22 a month. More than 60 of my shorts are voiced for free, and the voices are neural quality. This post is that setup.
The free part nobody mentions
Microsoft Edge, the browser, ships with a read-aloud feature backed by
neural voices. Those same voices are reachable programmatically, with no
API key and no account, through an open-source library (edge-tts on
Python). Hundreds of voices across dozens of languages, including the
natural-sounding Korean voices my channels run on.
My pipeline feeds it the script, scene by scene, and gets back narration audio files. Cost per video: zero. Cost across 60+ videos: still zero.
The trick that matters more than the voice: auto-length
Shorts live or die by length. My format targets 15-20 seconds, and scripts never come out at exactly the right length.
So the pipeline measures the narration and adjusts the speaking rate automatically until the total lands inside the window. A slightly long script gets read a touch faster; a short one, slower. No re-writing, no manual trimming. Combined with per-scene audio measurement, the subtitles stay in sync no matter what rate it picked.
This one feature is why the system produces consistent-length videos without me thinking about duration at all.
What you can use today
Three commands and you have your first free narration:
pip install edge-tts
edge-tts --list-voices
edge-tts --text "This voice costs nothing" --voice en-US-AriaNeural --write-media test.mp3
The second command prints every available voice. Pick one for your language, swap it into the third command, and you have narration audio ready to drop on any timeline. That is the entire barrier to entry.
Honest limits
| Question | Honest answer |
|---|---|
| Is it as expressive as premium AI voices? | No. Top-tier paid voices act; these read well. |
| Good enough for faceless shorts? | Yes. Retention on my best short is 74.9%. |
| Emotional storytelling, audiobooks? | Use a paid voice. This is the wrong tool. |
| Terms of service? | It rides an undocumented endpoint. Like the CapCut draft trick, it can change under you. That is the price of free. |
I have not run an A/B test of free versus paid voices on the same video yet. When I do, the numbers will be posted here, whichever way they fall.
What it saves
A paid TTS plan at $11-22 a month is $132-264 a year. For a channel already earning, that is nothing. For someone testing whether their channel idea works at all, that is the difference between trying and not trying. My whole shorts pipeline costs $0 in subscriptions, which means every experiment I run is free to fail.
Next up in the series: the auto subtitle system, and my ongoing fight with a 59.9% first-second swipe-away rate.
Subscribe below to get it. Here are the receipts.