Pick one voice and keep it for the next 200 videos. Our position: unless the voice is genuinely yours, design it rather than clone one. A designed voice has no person attached to it, so there's no consent to chase and no one who can later withdraw it, and that matters more than any quality difference between the two.
That's the short answer for a cloned voice for a faceless channel versus a designed one. The rest of this post covers the cases where we'd clone anyway, how each option holds up over hundreds of uploads, and what each one costs in points, so you can disagree with us on the facts.
Voice design vs voice cloning: what each one is
Voice cloning starts from a 15 to 30 second recording of a real voice, with 30 seconds recommended. You confirm that the voice is yours or that you have permission, and the clone is ready in under a minute. Voice design starts from words instead. You write an English description of age, accent, pace and tone, compare the preview takes, and save the one you like. See the voice design page for the flow.
Both options end in the same place: a saved voice in your library that you pick from the same menu as any ready-made voice. After that, the script and the points work identically for both.
| Question | Designed voice | Cloned voice |
|---|---|---|
| Whose voice is it? | Nobody's. It's original. | A real person's, with their consent. |
| Starting material | An English description | 15 to 30 seconds of clean speech |
| Cost to create | Preview takes use points | No points on any plan |
| Slots (Free, Lite, Pro, Max) | 1, 1, 5, 40 | 1, 1, 10, 200 |
| Best for | A channel brand that isn't a person | Sounding like you at scale |
Our pick for most faceless channels
A faceless channel is a brand, not a personality, and a brand voice doesn't need to belong to anyone. Design it. You'll end up with something distinctive that no other channel has, and no one's phone recording sits underneath all 200 videos. If a narrator hired for 20 videos leaves, a clone of their voice becomes a question you didn't want. A designed voice never leaves.
Design has one real drawback, and it's control. You're steering with words, so the first takes may miss. Expect to describe, listen and adjust a few times, and read how to describe a voice for AI before you start, because vague adjectives give vague voices. A designed voice also won't sound like you, which is the entire point when you don't want to be on the channel.
Keeping a consistent voice across hundreds of videos
Consistency is the same for both, and it comes from saving the voice, not from the method you used to make it. Once a design or a clone is saved, every script you generate with it uses that voice. Treat the saved voice as the asset. The description you wrote is just the recipe, and running it again gives you new takes, not the same voice.
So save the winner immediately, name it with the channel, and don't re-describe it. For team channels, the same saved voice can narrate whatever script anyone writes, which keeps the sound steady when editors change. You'll still hear the natural variation of any spoken read from one generation to the next. A little is normal; a lot usually means the script is the issue, and shorter sentences settle it.
If you're after a library voice instead of a custom one, a calm narrator such as this calm British male voice does the same job with no setup. The trade-off is that other channels can pick it too.
Changing your mind later is possible but visible. Videos you've already published keep the old voice, so a switch at video 60 gives you a channel with two sound identities, and regular viewers will notice within a single intro. That's the strongest argument for the audition below: decide at video 1, not video 60, when swapping the voice means re-narrating a back catalogue or living with the seam.
Who owns the consent, and who owns the voice
This is where the two options really part ways. When you clone your own voice, the consent question is easy: it's yours. When you clone a hired narrator, you need their written agreement, it should say explicitly that it covers AI voice use, and it should say for how long. We're not lawyers and this isn't legal advice. It is the paperwork we'd want before 200 videos depended on someone else's voice.
A designed voice sidesteps all of it, and you can use a saved voice commercially on a paid plan. Free-plan output is for personal, non-commercial testing. The commercial rights explained plainly post covers what each plan allows.
What each option costs
Creating a clone costs no points. Designing a voice does, because preview takes are generated like any other audio. If your test line is 200 characters, each take costs ⌈200 ÷ 17⌉ = 12 points, so 3 takes is 36 points, and a few rounds of tweaking will run to a couple of hundred. That's real but small next to a month of narration.
The narration itself costs the same either way: ⌈characters ÷ 17⌉ points, about 1 point per second of audio. An 8-minute video is roughly 480 points. Lite's 3,000 monthly points cover about 6 of them, and Pro's 12,000 cover about 25. Plan around retakes too, since every regeneration is billed again.
Audition before you commit to 200 videos
Whichever route you take, don't judge a voice on its welcome line. Run the same 3 scripts through every candidate: a 20-second intro, a 1-minute stretch heavy with numbers and names, and one emotional line. The intro tells you if the voice has a personality. The numbers stretch tells you whether it stumbles on the dull parts that fill most videos. The emotional line shows you where its range ends.
Do it back to back, with headphones, on a different day than the one you made the voice. Then ask a blunt question: could you listen to this for 8 minutes at 1.25 times speed without wincing? A voice that passes that test has earned a place on a channel. A voice that fails it will cost you viewers long before it costs you points.
When we'd clone instead
There are 3 good reasons. You want viewers to recognize you, and you're comfortable being the voice but don't want to record every script. You're a solo creator who currently records live and is tired of re-takes. Or a narrator you work with has agreed, in writing, to a clone and is paid for it.
- It's your own voice. Consent is simple, and the channel is built on you.
- You'd record anyway. The clone saves your throat and your evenings, not your identity.
- There's a signed agreement. Anyone else's voice needs one that names AI use and a term.
- You have the slots. Free and Lite hold 1 clone, so a second voice means a plan change.
For the channel-building side of this, our AI voice clone for faceless YouTube channels page walks through cloning a brand voice.
Script to finished narration is a separate job. Generate faceless YouTube video voice covers it.
Frequently asked questions
Should a faceless YouTube channel use a designed or a cloned voice?
For most, a designed voice. It's original, has no person attached, and needs no consent paperwork. Clone instead when the voice is your own or a narrator has agreed in writing to AI use.
Will a designed voice stay the same across videos?
Yes, once you save it. Every script generated with the saved voice uses that voice. Re-running the description gives new takes, so save the one you like straight away.
How much does it cost to create each kind of voice?
Creating a clone costs no points. Voice design uses points for its preview takes, as with any generation. Narration then costs the same for both, at ⌈characters ÷ 17⌉ points in the app.
Can I use a saved voice commercially on a faceless channel?
On a paid plan, yes. Free-plan output is for personal, non-commercial testing, so move to a paid plan before monetizing the channel.