Voice Banking: A Compassionate, Practical Guide
What voice banking is, who it helps, when to start after an ALS or MND diagnosis, what it realistically costs, and how to keep not just the voice but the person behind it.
Voice banking means recording your voice while your speech is still strong, so a personal synthetic voice can be built for you to use later. It matters most after an ALS or MND diagnosis, before a laryngectomy, or for a parent whose voice you want to keep. Start early: stronger recordings make a truer voice.
If you are reading this soon after a diagnosis, take a breath. Nothing on this page needs to be done today, and none of it commits you to anything. Voice banking is simply a way of keeping choices open. This guide explains what it is, who it helps, when to start, what it costs, and how to do it well. It is general information, not medical advice: your speech and language therapist knows your situation and should guide the timing.
This guide is published by Afterlife AI™. We make a voice and legacy preservation product ourselves, so treat us as an interested party: we have kept every factual claim checkable against the official sources listed at the end, and we point you to free and clinical tools before our own.
On this page:
What voice banking is (and what it is not)
Voice banking is the process of recording samples of your speech, usually scripted sentences, so software can build a synthetic voice that sounds like you. That voice is then loaded onto a communication aid (a speech-generating device, often called an AAC device) or an app. Later, if speaking becomes hard, you type or select words and the device says them out loud in a voice that is recognisably yours, rather than a generic computer voice.
It sits alongside a second, equally important practice: message banking. Message banking means recording real phrases in your own voice, exactly as you say them: the way you say each family member’s name, your standard jokes, "goodnight, sleep tight", "I’m proud of you". These are kept as actual recordings, not synthesised, so the warmth, timing and laughter are all preserved. The approach was pioneered by speech pathologist John Costello at Boston Children’s Hospital, and most clinicians now recommend doing both: a synthetic voice for everyday conversation, real recordings for the phrases that carry the most love.
Two things voice banking is not. It is not only for people who are dying: it is planning, the same way a will is, and many people bank a voice they never need. And it is not the shady end of voice cloning. Done properly it is consent-based from the first minute, because you record your own voice, for your own future use, on terms you choose.
Who voice banking is for
People living with ALS or MND. Amyotrophic lateral sclerosis, called motor neurone disease in the UK, Ireland and Australia, is the diagnosis that drives most voice banking. Many people with ALS or MND will notice changes to their speech at some point, and in bulbar-onset disease speech and swallowing are affected first, sometimes before anything else. That is why clinics raise voice banking soon after diagnosis: not to alarm anyone, but because the recordings are best made while speech is easy.
People facing a laryngectomy. A laryngectomy is surgery that removes the voice box, usually to treat laryngeal or other head and neck cancers. Unlike a progressive illness, the timeline here is fixed: recordings must be made before the operation. If surgery is scheduled, even a short recording session in the days beforehand preserves something no one can recreate afterwards.
People with other progressive conditions. Speech can change with Parkinson’s disease, multiple sclerosis, Huntington’s disease and some other neurological conditions. The progression is usually slower than ALS, which makes it tempting to wait. The recordings, though, only get harder to make.
Aging parents, and anyone at all. You do not need a diagnosis. Many families simply want to keep the voice of a mother or father while it is strong: the accent, the phrasing, the particular way they tell a story. If that is you, our guide to preserving a parent’s voice walks through it gently, step by step.
When to start: early, while your speech is strong
Every clinical source says the same thing: start early. The MND Association and the ALS Association both encourage people to bank their voice soon after diagnosis, while speech and breath support are strongest, because the quality of the recordings sets the ceiling on how natural the synthetic voice will sound. Early recording also means shorter, easier sessions: a strong voice can finish scripts in a fraction of the time a tired one can.
Starting early does not mean rushing. Modern tools need far less audio than they used to, and sessions can be spread over days or weeks. Fifteen minutes here and there is enough to make real progress, and everything you record is banked the moment you record it. You lose nothing by pausing; you only lose options by never starting.
If speech has already changed, do not assume the door is closed. Some services can strengthen a voice built from affected speech, and some can work from older material: home videos, voicemails, wedding speeches. Gather what exists before deciding anything.
And if you want somewhere gentle to begin today, you can start free with Afterlife AI: 50 memories, no card, with voice recording included in the free build. Recording your voice and your life story can happen in the same sitting, one prompt at a time.
How voice banking works, step by step
The details vary by tool, but the shape of the process is the same everywhere.
Talk to your care team first. Ask your speech and language therapist (SLT in the UK, speech-language pathologist or SLP in the US) which tools they support and whether funding is available. They will also check practical things like breath support and session length.
Choose your tools. Most families end up doing two or three things at once: a clinical voice bank for the device voice, message banking for real phrases, and a legacy layer if they want the stories and the person preserved too. The comparison table below covers the main options.
Set up a quiet space. A small room with soft furnishings, no fans or appliances running, and a phone or USB microphone held at a consistent distance. Quiet matters more than expensive equipment.
Record in short sessions. Fifteen to twenty minutes at a time, ideally when energy is highest (for many people that is morning). Stop before you tire; a strained voice today becomes a strained synthetic voice forever.
Read the scripts naturally. Speak the sentences the way you talk across a kitchen table, not the way an announcer reads the news. The software is learning you, so give it the real you.
Build and test the voice. Once enough audio is in, the service generates your synthetic voice. Listen to it on the actual device or app you plan to use, and re-record weak sections if the tool allows it.
Save everything, and decide who can reach it. Keep your original recordings in two places (a computer plus a cloud drive), uncompressed where possible. Then decide, deliberately, who should be able to access the voice and recordings later. That last decision is the one most tools skip; we come back to it below.
Your options compared: clinical tools and consumer apps
There are two broad families of tools. Clinical voice banking services are built for people who will use the voice on a communication device, and are the ones your speech therapist will know. Consumer voice apps are built for speed and natural sound, and a legacy service like ours is built for what remains around the voice. Prices below are as published in mid-2026 and change; always confirm on the official sites, which are linked in the sources.
Option | Type | What you record | Typical cost (mid-2026) | Best suited to |
|---|---|---|---|---|
Apple Personal Voice | Built into iPhone, iPad and Mac (iOS 17 or later) | About 150 phrases, roughly 15 minutes | Free with a compatible Apple device | A safeguard anyone can complete this week; pairs with Live Speech |
Acapela my-own-voice | Clinical, integrates with AAC devices | From around 50 sentences; more improves quality | Recording is free; the finished voice is licensed, often through your device provider. Confirm current pricing | People who will use a speech-generating device day to day |
SpeakUnique | Clinical, UK-based, AAC-friendly | Around 150 to 300 sentences; can also work with changed speech or old recordings | One-off build fee; often funded for people living with MND in the UK. Confirm current pricing | Early speech change, and UK and Ireland families |
ModelTalker | Non-profit research service (US) | Several hundred up to about 1,600 sentences | Low cost, donation-supported. Confirm current pricing | Budget-conscious US families who can record a lot |
Consumer AI voice apps | Consumer voice cloning | A few minutes of clear audio | Usually a monthly subscription; varies widely | A fast, natural-sounding voice; check consent, export and cancellation terms first |
Afterlife AI | Legacy preservation: Persona plus voice | Guided memories and stories, plus your voice, recorded with your consent | Free to build (50 memories, no card); listening to the preserved voice is part of paid plans from $14.99/mo | Preserving the person behind the voice: stories, values, the way you think |
These are not competitors so much as different jobs. A clinical voice bank gives you a voice that can say new sentences on a device across the kitchen table. Message banking keeps the true sound of the phrases you actually say. A consumer app gets you a natural-sounding copy quickly; our guides to the best app to preserve your voice and how to clone your voice safely compare those in detail. Many families sensibly do more than one.
What makes a good recording
Whatever tool you choose, the recordings decide everything. The good news is that great recordings need care, not money.
Quiet beats fancy equipment. A silent room with a phone at a steady distance beats an expensive microphone next to a humming fridge. Turn off fans, close windows, silence notifications.
Keep the microphone distance consistent. Around 15 to 20 centimetres, the same in every session, so the voice sounds uniform.
Speak like yourself. Conversational pace, natural pauses, your normal warmth. Resist the urge to perform; the goal is you, not a broadcaster.
Short sessions, water nearby. Fatigue changes a voice quickly, especially with MND. Two relaxed 15-minute sessions beat one exhausting hour.
Record when energy is highest. For many people that is the morning. Follow your own pattern, not a schedule.
Re-record rather than settle. If a sentence came out stiff or you stumbled, do it again. Every fix now is permanent quality later.
Keep the originals. Save uncompressed files (WAV where the tool allows) in two separate places. Originals can be reused by better technology in the future; a compressed copy cannot be uncompressed.
Message banking: the phrases only you can say
A synthetic voice can say anything, but it says everything the same way. Message banking exists because some phrases are not information, they are performances: the exact rise and fall of "come on then, dinner!", the laugh in the middle of a story that has been told a hundred times, a grandchild’s name said the way only one person says it.
Build a simple list and work through it in relaxed sessions. Families who have done this suggest including: each person’s name as you actually say it; your greetings and sign-offs; your sayings and catchphrases; "I love you" and "I’m proud of you"; a birthday message; a goodnight line for small children; and one story told all the way through, laughter included. Record them as real audio, keep the originals safe, and label the files clearly while you still remember the context of each one.
If you are helping a parent do this, be patient with the strange feeling of it. Almost everyone feels self-conscious for the first ten minutes and glad for the rest of their life.
Beyond the voice: preserving the person
Here is the honest limit of everything above: a banked voice preserves how someone sounds. It does not preserve what they would say. The stories, the values, the advice, the way they think through a problem, the things they would want a grandchild to know at 18: none of that lives in a voice file.
That is the part Afterlife AI™ is built for, and to be transparent, it is our product. You build a Persona while you are alive, in your own words: guided prompts help you record memories, stories and beliefs, and your voice is preserved alongside them with your explicit consent. That consent expressly covers playback after you are gone. Executor Lock™ then lets you decide, in advance, exactly what loved ones can reach later, and seals those choices so they are never changed after death. Nothing autoplays in grief contexts; a family member always chooses to tap before hearing anything. Afterlife AI is an Australian company, your content is hosted and stored in Australia, and your voice is treated as sensitive information under Australian privacy law.
To be clear about what it is not: Afterlife AI is not a communication aid and does not replace the clinical tools above. If you are living with ALS or MND, bank your voice with a clinical service for daily communication, and use Afterlife AI for the layer no device covers: the person. Building is free (50 memories, no card, and the free build never expires), voice recording is included, and listening to the preserved voice is part of the paid plans on our pricing page. Family inherits the time you have paid for.
Frequently asked questions
Short, direct answers to the questions families ask most. Medical decisions belong with your care team; this is general information.
What is the difference between voice banking and message banking?
Voice banking builds a synthetic version of your voice from scripted recordings, so a device can later say anything you type in a voice that sounds like you. Message banking saves real recordings of your actual phrases, exactly as you said them: names, sayings, laughter, "I love you". They meet different needs, and most speech therapists suggest doing both. The synthetic voice covers everyday communication; the real recordings keep the true sound of you.
When should I start voice banking after an ALS or MND diagnosis?
As early as you comfortably can, while your speech and breath support are strong. Speech changes arrive at different times for different people, and in bulbar-onset disease speech is often affected first. Starting early keeps every option open, and recording something never obliges you to use a device later. Even a single 15-minute session with a free tool such as Apple Personal Voice is a meaningful safeguard you can do this week.
Can I still bank my voice if my speech has already changed?
Often, yes. Some clinical services, such as SpeakUnique, can build or repair a voice from speech that has already changed, or from older recordings like home videos and voicemails. The result depends on the quality and amount of material available, so gather everything you have. A speech and language therapist can tell you what is realistic in your situation. Do not rule yourself out without asking.
How long does voice banking take?
Far less time than it used to. Apple Personal Voice needs about 15 minutes of phrases. Modern clinical tools typically ask for between 50 and a few hundred sentences, recorded in short sessions over days or weeks at your own pace. Older research systems needed 1,600 or more sentences, which is where the reputation for being exhausting came from. Message banking has no fixed length: you simply record the phrases that matter, whenever you can.
How much does voice banking cost?
It can be free. Apple Personal Voice costs nothing on a compatible iPhone, iPad or Mac. Clinical services often let you record for free and then charge a licence or one-off fee for the finished voice, and charities frequently cover that for people with ALS or MND. With Afterlife AI you can start free: 50 memories, no card, with voice recording included in the free build. Listening to the preserved voice is part of paid plans, with Legacy at $14.99 a month.
Is there financial help for people with ALS or MND?
Usually, yes. In the United States, nonprofits including Team Gleason and the ALS Association have programs that help cover voice preservation and communication technology for people living with ALS. In the UK, the MND Association has offered support toward voice banking costs, and NHS speech and language therapy teams can advise on funded options. Programs and criteria change, so ask your clinic or local association what is current before paying for anything yourself.
What happens to my banked voice after I die?
It depends entirely on the provider, so ask directly: who can access the voice, under what terms, and for how long. Many tools were designed for daily communication and say little about afterwards. Afterlife AI is explicit about this. You decide while alive, your consent expressly covers playback after you are gone, and Executor Lock seals those choices so access never drifts beyond what you agreed to. Nothing autoplays; a family member always chooses to tap before hearing anything.
Is voice banking safe, or could my voice be misused?
Reputable services build your voice only with your explicit consent and publish terms covering storage, deletion and export: read them before you record. Consent is what separates voice banking from impersonation scams, which use stolen audio without permission. Our guide to whether voice cloning is safe covers the risks and protections in detail. With Afterlife AI, your voice is treated as sensitive information under Australian privacy law and your content is hosted and stored in Australia.