On this page
Twenty-nine videos averaging just under five minutes is 142 minutes of narration in one month. On hosted pricing that exceeds the 121 minutes included at $11 and lands you on the $99 tier. On a voice model that runs on the laptop it is a 115 MB download and nothing per minute.
The difference between those two is real but it is not the one people assume. It is not intelligibility — it is expressiveness, and whether your content needs any.
What do the hosted tiers actually cost?
| Plan | Price | Credits | Roughly |
|---|---|---|---|
| Free | $0 | 10,000 | ~10 minutes, no commercial licence |
| Starter | $6 | 30,000 | ~30 minutes |
| Creator | $11 | 121,000 | ~121 minutes |
| Pro | $99 | 600,000 | ~600 minutes |
From ElevenLabs' published pricing, checked 21 October 2026, at roughly one credit per character.
The row that matters most is the first, and not for its size. The free tier has no commercial licence. If the video is going on a monetised channel or in front of something you sell, the free plan is not a cheaper version of the paid one — it is a different licence. The real entry price is $6.
How much narration does a series actually need?
More than the estimate in your head.
| Output | Minutes of audio |
|---|---|
| One five-minute video | 5 |
| One a week for a month | ~20 |
| 29 videos at ~4.9 minutes | 142 |
Twenty videos a month is an unremarkable publishing rate for anyone doing this seriously, and it is already past the $11 tier. The pricing is built around occasional use, and a content pipeline is not occasional use.
Then there is the part that is not in any table: regeneration. Every rewritten sentence re-spends the characters. If the first version of a script is the one that ships, the numbers above hold. If you edit — and you should — the real consumption is somewhere between 1.5 and 3 times the finished length.
The whole list
67 tools, what each costs and when you actually need it. Updated twice a year.
What does a local model give up?
Prosody, mostly. Not clarity.
A modern local voice model is a file of around 100 MB that runs on an ordinary laptop with no API call, no per-character cost and no monthly ceiling. The 29 videos on this project were narrated that way.
| Hosted | Local | |
|---|---|---|
| Cost per minute | metered | zero |
| Commercial use | from $6/month | yours |
| Regeneration | costs again | free |
| Emotional range | wide | narrow |
| Emphasis and pacing | controllable | mostly flat |
| Offline | no | yes |
The honest summary: a local voice reads correctly and does not act. It will not land a joke, lean on a word, or sound moved. For an explainer where the information is the point, that gap is smaller than its price. For a story, it is the whole thing.
Where does flat narration actually hurt?
In three specific places, and you can write around all of them.
Lists. A flat voice reading seven items in a row is a lullaby. Break the list into sentences, or put it on screen and say less.
Numbers. "Fifty-five per cent" delivered without emphasis slides past. Give the number its own short sentence.
The turn. Every explainer has one moment where the point arrives. A hosted voice can carry that with delivery; a local one cannot, so it has to be carried by the writing or by what is on screen at that second.
That is the real trade: a flat voice makes you write better, because nothing else is going to do the work.
When is paying clearly right?
When the voice is part of the product rather than a delivery mechanism.
- A recognisable narrator. If people are coming back partly for how it sounds, that is a brand asset and $11 a month is nothing.
- Anything with character or dialogue. Flat delivery is fatal here, not merely weaker.
- Short, high-stakes pieces. A 60-second ad where every word is weighted is exactly where 30 minutes a month is plenty and the quality gap is most visible.
And when is local clearly right? High volume, informational content, where you will regenerate a lot and nobody is listening for the performance. Which describes most explainer and tutorial output, including this project's.
What about the editing tools that include a voice?
They are a different purchase. An editor with text-to-speech bundled is priced as an editor, and the voice is a feature rather than the product — usually with a smaller allowance and less control. Can you edit video without a monthly subscription? covers the pipeline side of this, and what AI tools actually cost per month puts the whole stack together.
Where these numbers came from
ElevenLabs' published pricing page, checked 21 October 2026, in US dollars, with the stated approximate minutes per tier. The 142 minutes is measured from the 29 finished videos on this project — 29 files, 2.07 GB, averaging 4.9 minutes.
Credit allowances change and generally grow. The structural points do not: the free tier is a different licence, regeneration costs again, and a content pipeline consumes far more than the pricing page's examples assume.
The AI Persona Playbook covers the production side of running a channel like this.
Tools in this piece
| Tool | What it does | Price |
|---|---|---|
| ElevenLabs | Voice generation and voice-overs that do not sound like a robot. | Free tier, then ~$5–22/mo |
| Descript | Edit video by editing the transcript. | Free tier, then ~$15–30/mo |
| CapCut | Fast shorts and vertical editing, with captions. | Free, Pro ~€10/mo |
Every tool has its own page with the price, who should skip it and what to check before paying. Some links are affiliate links.
Questions people ask
How much does AI text to speech cost?
On ElevenLabs' published pricing, checked 21 October 2026: free gives 10,000 credits, roughly ten minutes of audio, with no commercial licence. Starter is $6 a month for about 30 minutes and adds the commercial licence. Creator is $11 for about 121 minutes and Pro is $99 for about 600.
Can you use the free tier for monetised videos?
No. The free plan does not include a commercial licence. For anyone publishing to a monetised channel or selling a product, the $6 tier is the real entry point, not the free one.
Is a local text-to-speech model free?
Yes, once you have the model file. A voice model is a download of around 100 MB that runs on an ordinary laptop with no per-character cost and no monthly limit. What you give up is expressiveness, not intelligibility.
How many minutes of narration does a video series need?
More than people estimate. Twenty-nine videos averaging five minutes is 142 minutes of audio, which exceeds the 121 minutes on the $11 tier in a single month.
When is paying for a hosted voice worth it?
When the voice is the product — a narrator people are listening to for its own sake, or a brand voice you reuse. For explainer content where the words carry the value, the gap between hosted and local is much smaller than the price gap.
Does regenerating audio cost you twice?
Yes, and this is the hidden cost. Every rewrite of a script re-spends the characters. A local model makes iteration free, which quietly changes how willing you are to fix a bad sentence.