Voice cloning has one legitimate use that covers almost everything worth doing with it, and one absolute prohibition. The legitimate use: cloning your own voice, so you can produce narration without recording every time. The prohibition: cloning anyone else's without documented consent for the specific use.
That second one is not a grey area, it is not a matter of platform policy, and it is where this technology causes real harm.
TL;DR
- Your own voice: legitimate and genuinely useful for narration at volume
- Anyone else's: needs documented, specific consent. Public figures: never
- Free tiers typically offer a limited clone and a small monthly character allowance
- On 8frame: ElevenLabs TTS at 13 credits per 1,000 characters. We have library voices rather than a cloning tool
- Quality depends on your source recording far more than on the tool
The consent rule, stated once and properly
Cloning a voice requires the person's permission, and the permission has to cover the specific use.
Someone who let you clone their voice for an internal training module did not agree to a product advertisement. Someone who appears in your podcast did not agree to be synthesised at all.
Never a public figure. Not as a joke, not as a demo, not as a proof of concept. Voice is one of the two biometric identifiers people cannot change, and synthetic speech attributed to a real person is the mechanism behind the most damaging misuse of this technology: fraud against their family, fabricated statements, and financial scams.
Serious platforms enforce this with a recorded verification step where the speaker states consent aloud. That is not bureaucracy; it is the thing that makes the feature defensible.
Cloning your own voice, which is the useful case
What it is good for:
- Narration at volume. A course, a documentation series, a set of tutorials, where recording every update is impractical.
- Corrections. Fixing one sentence in a finished video without rerecording the session or matching room tone.
- Consistency across a long project recorded over months in different rooms.
- Localisation of your own material, with a native speaker checking every language.
What it is not good for: anything where the point is that you showed up. A founder video, a personal message, an apology. The audience is responding to presence, and presence is what a clone removes.
Recording a good source
Quality here depends on your recording far more than on the tool:
A quiet room. Not silent, quiet. Soft furnishings, no fan, no fridge, no traffic.
Consistent distance. A fixed position from the microphone throughout. Moving closer and further teaches the model inconsistency.
Read varied material. Not one flat paragraph: include questions, emphasis, longer and shorter sentences. A clone trained on monotone reads everything monotone.
Normal pace. Reading unnaturally slowly for clarity produces a clone that always sounds like it is reading.
Enough material, per the tool's guidance, and a bit more.
What it costs
Canon 8frame prices, a credit is $0.01 at pack rate:
| Task | Route | Credits |
|---|---|---|
| Narration, 1,000 characters | ElevenLabs TTS | 13 |
| A 500-word script | ElevenLabs TTS | ~40 |
| A 90,000-word audiobook | ElevenLabs TTS | ~6,500 |
| Sound effect | ElevenLabs SFX | 5 |
| Music bed, 30 seconds | Lyria 2 | 13 |
To be clear about scope: 8frame offers text to speech with library voices. It is not a voice cloning platform, and we are not going to imply otherwise. Cloning happens on a dedicated service, and the generated audio can then be used in your work here.
What free tiers give you
Typically: one custom voice, a monthly character allowance in the low tens of thousands, and non-commercial terms. Enough to evaluate whether a clone of your voice is good enough for your purpose, which is the right question to answer before paying.
Check two things in the terms before uploading anything: whether your voice recordings are used for training, and whether there is a deletion path. Your voice is biometric data and free tiers are more likely than paid ones to fund themselves with it.
The disclosure question
Even with your own voice and full rights, there is a judgement call about telling people.
Where disclosure clearly matters: anything where the audience believes they are hearing you speak to them directly, and anything where the content is personal, emotional or persuasive about you.
Where it clearly does not: documentation narration, training modules, and informational material where nobody assumes a live performance.
The reasonable default is to disclose when the audience would be surprised to learn it. That test handles most cases.
FAQ
Is voice cloning legal? Cloning your own voice, generally yes. Cloning someone else's without consent creates liability in most jurisdictions and is prohibited by every serious platform.
Does 8frame do voice cloning? No. It offers text to speech with library voices at 13 credits per 1,000 characters.
How much source audio do I need? Follow the tool's guidance and provide a bit more, with varied delivery rather than a flat read.
Should I disclose a cloned voice? When the audience would be surprised to learn it, yes.
Clone your own voice, never anyone else's. The 8frame canvas is free and unlimited, and generation is paid from $19/month.