Fish Audio Voice Cloning From a 15-Second Clip

What voice cloning means here

Cloning in Fish Audio takes a short recording and builds a voice you can reuse. Fifteen seconds is enough: the model picks up tone, pitch, and speaking style from the sample, and from then on you generate new lines in that voice instead of recording them.

That is the practical difference between cloning and simply picking a library voice. A library voice gives you a known, consistent narrator. A clone gives you a specific voice: the one in your existing footage, the one a client approved, or the one your audience already recognises. For a series, that consistency is usually the whole point.

The sample does not need to be studio quality. It needs to be clean and representative: one speaker, no music, no overlapping voices, and a natural reading pace rather than a performance.

How to record a sample that works

Most disappointing clones trace back to the recording, not the model. A few habits make the difference:

• Record in a quiet room, close enough to the microphone that the voice is clearly louder than the room.
• Read in your normal speaking tone for the project. A whispering sample produces whispered lines.
• Keep the pace even. Fifteen seconds of natural speech is more useful than five seconds of acting.
• Avoid background music, notification sounds, and second speakers.
• Leave a little silence at the start and end so the clip boundaries are obvious.

If the source material already exists, an interview clip or a clean voiceover segment usually works well. If you are starting from scratch, reading two or three sentences aloud is enough to reach the fifteen seconds.

What you can do with a cloned voice

Once the voice exists, it behaves like any other voice in the workspace. You can:

• Generate new lines from text, with emotion tags controlling delivery where the script changes mood.
• Keep a series consistent by reusing one voice across episodes, chapters, or ad variants.
• Use the voice across 30+ languages, so the same character speaks to every market.
• Compare takes in the same way you would with a library voice, revising the script rather than the voice.

Cloning sits alongside the rest of the engine rather than replacing it. If you only need a narrator, the voice library may already have one; if you need speech without a cloning step at all, start with text to speech.

Consent, rights, and commercial use

Clone a voice you have the right to use. That normally means your own voice, a voice you have recorded with documented permission, or a voice you have licensed. Cloning someone without consent can break platform rules, and in many places it breaks the law.

For output, the same commercial rules apply as everywhere else in the product: personal projects are covered by any pack, and commercial usage rights come with the Creator and Studio packs. Every generated file is worth reviewing before it goes out, which is what the site disclaimer asks for: listen to what you are publishing, and make sure you have the rights to publish it.

If you are producing for a client, keep the sample and the written permission together. It takes a minute at handover and saves a difficult conversation later.

Frequently asked questions

How long does a sample need to be?

Fifteen seconds is enough. Longer samples can help with unusual voices or strong accents, but they are not required, and a clean short clip beats a noisy long one.

Can I clone a voice and use it in another language?

Yes. A cloned voice works across the same 30+ languages as the rest of the platform, so you can keep one voice consistent in every locale.

Is cloning included in the $9 Starter pack?

Yes. Voice cloning is part of every pack, including the $9 Starter pack. Commercial usage rights, however, start with the Creator pack on the pricing page.