Custom Voice: make a voice from your own recording, and keep it yours
Published · Last updated
Part of the guide: Choosing a Voice Guide
A custom voice in Syllivo is not a preset you pick from a list. It is a voice built from a short recording you supply yourself, so the result sounds like the person in that recording rather than like a numbered preset. If you have not picked a voice yet, start with choosing a voice by characteristic, and when you are ready to try your own recording, open Studio and work from there.
What Custom Voice is, and what it is not
Custom Voice has no preset voices at all. You supply a reference recording, give the profile a name you will recognise, and generation uses that recording's voice. It does not invent a voice for you, and it does not turn someone else's voice into yours: the voice you get is only as good, and only as legitimate, as the recording you put in.
Like the rest of Syllivo, generation runs on your device. Your script and your audio are not sent to a voice-generation server to make the speech. There is also no artificial character or per-generation quota: the practical limits are your device's speed and storage, not a monthly allowance.
A good reference recording
The recording rules are short, and they matter more than any other choice you make:
- Between 3 and 30 seconds long.
- A single speaker, speaking naturally, with no second voice in the background.
- A quiet recording: no music, no television, no traffic, no echoey room if you can avoid it.
- Your own voice, or a voice you have the right to use. Syllivo's terms require rights over a Custom Voice reference, and impersonating someone else is not allowed.
A clean ten seconds of ordinary speech beats a long, noisy recording every time. Read a few normal sentences aloud rather than performing a character voice: the profile should capture how the speaker normally sounds, not how they sound at their most dramatic.
Download once, then test short
Custom Voice is the heaviest voice option in Syllivo, at about 1.7 GB, so its files take longer to download than the other voices and need more storage. Download them once, on a connection you trust, and they stay cached on your device for later generations.
Before running a long script, generate one short paragraph and listen to it properly. Check whether words are clear, whether the pacing sounds natural, and whether the voice still sounds like the speaker in your recording. If the test sounds wrong, fix the recording first: re-record in a quieter room, with one speaker only, before trying a different script or a longer passage.
Keep the original recording safe
Keep the original recording file somewhere safe, outside the app, under a name you will find again. If you do not retain the source audio, you will need to re-create the profile before generating again, so the recording is worth keeping even after the profile works. The same applies to the profile name: pick a name that tells you whose voice it is and what it is for, because a profile called after a person is easier to find months later than one called after a test.
If it still sounds off
Most problems with a custom voice trace back to the recording, not to the script. Work through this order: first re-record cleaner and shorter, then test a short paragraph again, and only then try rewriting the script with shorter sentences and cleaner punctuation. If a passage still sounds wrong after a clean re-recording and a rewrite, the honest answer may be that this particular recording is not a good reference, and a fresh one, recorded quietly and naturally, is the fix.
Frequently asked questions
How long should my reference recording be?
Between 3 and 30 seconds, with a single speaker in a quiet recording. A clean ten seconds of normal speech is a good starting point.
Can I use someone else's voice?
Only if you have the right to use it. Use your own voice, or a voice you have permission for. Syllivo's terms require rights over a Custom Voice reference, and impersonation is not allowed.
Why does Custom Voice take longer to set up than other voices?
Its files are the heaviest voice option in Syllivo, at about 1.7 GB, so the first download takes longer and needs more storage. After that first download, the files stay cached on your device.
What happens if I lose the original recording?
If you do not retain the source audio, you will need to re-create the profile before generating again. Keep the original recording file somewhere safe, under a name you will find again.