From the recordings your family already has

Keep Their Memory Alive in the Songs They Loved

Most families have more of a voice than they realise — a voicemail nobody deleted, a wedding video, a birthday song recorded on a phone. My Voice Studio can learn a singing voice from permitted recordings like those, and then sing the songs that person loved. It will not bring anyone back. It can keep the sound of them in the house.

Start with My Voice Studio How voice training works

My Voice Studio

Train your permitted voice once, then reuse it for covers

Private model

Step 1

Train the voice

Add clean recordings of one voice you own or have permission to use.

YouTube Upload Record

Step 2

Create the cover

Choose a source performance. Its melody, words, timing, and phrasing stay in place.

Source performance

YouTube · upload · microphone

Pitch ±12

New vocal identity, same performance

Use only your own voice or a singer who has explicitly permitted the training and intended use.

Explore My Voice Studio

How a memorial song is made

Three steps, and the first one is usually the longest, because it happens on old phones and in family WhatsApp groups rather than in the app.

Gather the recordings

Voice notes, home videos, wedding tapes, anything where they are singing or speaking clearly. Use material your family holds and has the right to use.

Train the voice

Add the recordings to My Voice Studio, review the set, and let the background worker build one model from that one singer. The studio shows the limits and the credit cap before you confirm.

Choose the songs they loved

Pick the tracks that meant something — the bhajan, the film song, the hymn — and create covers in the trained voice.

What you need, and what to expect

It is worth being honest about this before you start, because the material you have is the material you have.

Old recordings are rarely clean. Voxmin can prepare vocals from a full mix, but separation quality depends on the source, and background noise, music or other speakers all make the model less certain. A clear thirty-second phone recording of someone singing often gives a better result than a long, noisy one.

Sung material helps most

Speaking recordings add useful phonetic coverage, but a model meant to sing also needs sung examples across the notes you want to hear.

One person per model

Train one voice from one singer. Mixing two family members produces an ambiguous target and a less convincing result.

There is no magic amount

Coverage and cleanliness matter as much as duration. Start with the clearest material you have and stay inside the limits shown in My Voice Studio.

It is a likeness, not a resurrection

The model reproduces the character of a voice. It is recognisable and often deeply moving, and it is still a reconstruction.

Consent, rights and respect

Voxmin is built for voices you are permitted to use, and a memorial song is the case where that matters most. In practice that means the recordings belong to your family, and the people closest to the person — the surviving spouse, the children — are part of the decision rather than surprised by the result.

The songs themselves carry separate rights. Making a private recording to play at a family gathering is not the same as publishing it, and Voxmin does not grant rights to the original compositions.

Voice Preservation AI Lullaby Maker Train ai singing voice Ai song generator with my voice Ai singing voice generator

Start with My Voice Studio

Pay-as-you-go credits. Train the voice once, then create as many songs as you want.