Voice preservation
Voice Banking and Message Banking: How and When to Start
Voice banking and message banking record your voice now so a device can speak with it later, and both have to happen while your speech is still clear. Voice banking builds a synthetic voice from sentences you read. Message banking saves real phrases in your own voice, exactly as you said them. Start with a speech-language pathologist early, because waiting is what removes the option.
Show guidance for someone whose
Before you start
- A speech-language pathologist involved from the beginning, usually through a neurology or ALS clinic
- A quiet room with carpet and soft furnishings, which is what Team Gleason, a US nonprofit that provides technology and equipment to people with ALS, recommends to cut echo
- A computer and an internet connection, plus the browser the service asks you to use
- A headset microphone or a separate recorder, rather than the microphone built into a laptop
- A day when your voice is rested, and a plan to stop before it tires
- Your own list of the phrases that matter to you, written down before you sit down
The procedure
Bring in a speech-language pathologist first
Ask your clinic to connect you with a speech-language pathologist before you record anything at all.
The ALS Network, a US nonprofit that provides ALS care services in California and Hawaii, advises having a speech therapist measure your speech rate before or at the first sign of change, because that baseline is what tells the team when each communication step is due.
Choose your service before you record
Pick one service and confirm it produces a voice the phone, tablet, or communication device you plan to use can actually load.
The ALS Association, a US nonprofit that funds ALS research and provides care services through a nationwide network of ALS clinics, notes that Voice Keeper voices are made for specific platforms. Recording first and checking compatibility afterward can leave you holding a voice file your device will not accept.
Set up a quiet room
Turn off fans and air handling, close the windows, silence phones, and pick the softest room in the house.
Echo and background noise are recorded along with your voice, which is why Team Gleason asks for carpet and soft furnishings rather than a bare room.
Record a short test passage
Record one or two sentences, play them back, and listen for hum, hiss, or a level that distorts on louder words.
Finding a bad microphone after an hour of recording means spending that hour again, and energy is the resource you have least of.
Bank your own phrases first
Record the phrases you would want played back exactly as you say them, before you touch the read sentence set. The ALS Association gives “I love you, sweetheart” as its example.
The ALS Network describes message banking as preserving your cadence and unique delivery. Do that part while you are fresh, not at the end of a long session.
Work through the service’s sentence set
Read the sentences the service gives you at your normal pace and your normal volume.
Stop before your voice tires
Split the set across several short sessions instead of pushing through it in one sitting.
The ALS Network says voice banking is worth considering even if speech is slightly slurred, as long as you have the energy for the process. Energy is the limit, so protect it.
Keep your own copy of every raw recording
Save the original audio somewhere you control, such as an external drive plus a cloud folder, on top of whatever the service stores for you.
Boston Children’s Hospital calls it double dipping: using one round of message banking to also build a custom synthetic voice from 750 or more banked messages. That only works if the recordings still exist.
Install the voice on the device that will speak for you
Load the finished voice onto the phone, tablet, or communication device you will really use, and try it in an ordinary conversation.
Use caution
- Prices and recording requirements change. Every figure below comes from the ALS Association handout dated March 2025. Confirm current terms with the service itself before paying.
- Team Gleason asks you to apply for funding before making any payment, because it cannot reimburse costs paid before approval.
- The ALS Association notes ElevenLabs voices need an internet connection to use. Check what that means where your signal is weak.
- A banked voice is not a communication system by itself. It still needs a device and a way to select words.
Stop and get help
- Your voice tires, goes hoarse, or changes noticeably partway through a session
- Talking for a long stretch is leaving you short of breath
- Recordings keep getting rejected by the service for quality
- The finished voice does not sound like you and you are unsure whether to redo it
- You are about to pay for a service before checking whether Team Gleason will cover it
Print the nine-step checklistOne page, large type, keep it beside the computer you record on.
Three different things, often confused
Voice banking creates a custom synthesized voice from samples of your speech, in the ALS Association’s words. You read a set of sentences, software learns from them, and afterward the voice can say anything you type. It sounds like you in general, not in any particular moment.
Message banking is different. Team Gleason describes it as recording meaningful phrases in your own natural voice, including a laugh, and the recordings play back exactly as you made them. They are not recombined into new sentences. This is where the phrases that carry your personality survive.
Voice cloning is the route for people whose voice has already gone. Team Gleason describes it as creating a synthetic voice from audio that already exists, such as voicemails, videos, or other stored recordings. The ALS Association calls it an AI technology that produces a highly accurate digital replica of a person’s voice.
Most people are advised to do voice banking and message banking together. Cloning is the fallback, not the plan.
Which service, and what each one asks of you
The table below is the ALS Association’s own comparison from its voice preservation handout dated March 2025, which states the information was current at that date. Team Gleason says it works with the voice banking companies listed there, and will cover the cost of the software you choose if you live in the USA with a confirmed ALS diagnosis.
| Service | What you record | Time and cost as listed | Worth knowing |
|---|---|---|---|
| Acapela | 50 required messages plus custom ones, by app or website | 20 to 30 minutes. $999 lifetime or $99 annually | Has an option to build a voice from previous recordings. 17 languages. |
| SpeakUnique | 150 to 300 sentences for Voice Build | 45 to 90 minutes. Voice Build $235, Voice Repair $335, Voice Design $400 | Voice Repair and Voice Design use AI and a donor voice to recreate or replace lost speech. |
| Voice Keeper | Recordings made in an app | 10 minutes. $299 | Voices are made for specific platforms, iOS or Android. Confirm yours first. |
| ModelTalker | 300 recorded sentences for a Gen 3 voice | 60 to 120 minutes. $100 | Run by the Nemours Speech Research Laboratory. Custom inventory option for personal messages. |
| ElevenLabs | 30 or more minutes of existing quality audio | 30 minutes. Free | Cloning rather than banking. Needs an internet connection to use the created voice. |
| Apple Personal Voice | Recording on your own iPhone | Free, built in | The ALS Association recommends it if you already have an iPhone. Check Apple’s setup instructions for the current recording requirement. |
For message banking specifically, Boston Children’s Hospital points to free software at myMessageBanking, built around its own message banking process. Team Gleason also lists myMessageBanking and Acapela as message banking options. Ask your speech-language pathologist which combination your clinic already supports.
What you actually need at home
Less than people expect. The ALS Association says most companies require a computer, an internet connection, and a quality headset, and that the company will email you the equipment details once you are approved. Team Gleason adds that some tablets and phones work, that a specific browser such as Chrome or Edge may be needed, and that a headset or external recorder is optional but helpful.
For message banking, the ALS Association recommends a quality recorder or a phone app, naming AVR X and Easy Voice Recorder. Team Gleason has loaner equipment for voice preservation if buying a recorder is a problem.
How much recording is involved varies a lot by service. Team Gleason puts the range across providers at roughly 50 to 3,500 phrases.
ModelTalker’s own site describes a standard inventory of 1,600 sentences alongside a Gen3 inventory of 298 sentences designed to elicit expressive speech, and says staff can sometimes build usable voices from substantially fewer. ModelTalker is a US voice banking service run by the Nemours Speech Research Laboratory in Delaware. Ask the service how short its shortest usable set is before you commit.
What happens to the banked voice afterward
The finished voice is a file you install on something that speaks. SpeakUnique, a UK spin-out from the University of Edinburgh, says its voices work through its own app on Apple and Android devices, through Grid for iPad and Predictable, and on specialist Windows communication devices. ModelTalker says its voices work on Windows PCs, Macs, iOS devices, Android phones and tablets, and speech-generating devices from manufacturers including Tobii Dynavox.
Do not take any of that on trust for your own setup. Compatibility lists change, and a voice built for one platform may not move to another. Confirm with both the voice service and the device vendor, in writing, before you pay for either.
The device itself is the other half of the decision. Our guide to communication boards, speech generating devices and eye gaze covers how people select words once speaking is hard, and the communication and alerts equipment catalog shows what those devices look like in practice.
Voice preservation also sits alongside the paperwork decisions in life planning, and it is worth raising with the groups listed in our support network directories.
When this stops working
There comes a point where recording new material is no longer realistic, either because clarity has gone or because the effort costs too much. That is when cloning from audio you already have becomes the route.
Team Gleason describes building a synthetic voice from voicemails, videos, and other stored recordings, and lists free ElevenLabs licenses through Bridging Voice for people with ALS and MND. Bridging Voice is a US nonprofit that helps people with ALS keep communicating with assistive technology.
Voice cloning
ElevenLabs
A general purpose AI voice service that can build a synthetic voice from recordings you already have. It is not a purpose built assistive communication product, which cuts both ways.
- What it does
- Builds a synthetic voice from existing audio, so it is an option after speech has already changed
- Internet
- The ALS Association notes its voices need a connection to use
- Built for
- General AI voice work, not assistive communication
- Before you pay
- Confirm with your speech language pathologist and your device vendor that the finished voice will load onto the communication device you plan to use
Honest limitation: read this before you pay. There are free routes for people with ALS. Team Gleason lists free ElevenLabs licenses through Bridging Voice, and ElevenLabs runs an Impact Program offering free professional voice clones to people living with ALS. Check both before paying anything.
It is also not built for assistive communication, so nobody can promise the finished voice will run on your device. Confirm that with your speech language pathologist and the device vendor first.
We earn a commission if you sign up through this link. It does not change the price you pay, and it does not influence which services we cover. Check the free license route above first.
The ALS Association lists SpeakUnique Voice Repair and Voice Design as using AI and a donor voice to recreate or replace lost speech, and notes that Acapela and Voice Keeper both offer options built on previous recordings. Go looking through old phones and old videos before you decide it is too late.
Common questions
When should I start voice banking?
As early as you can. The ALS Association’s answer is to start as soon as possible, while your voice is the strongest. The ALS Network places banking before or at the first sign of speech change, and says it should still be considered if speech is slightly slurred, as long as you have the energy for it.
Is it too late if my speech has already changed?
Probably not, but the answer belongs to a speech-language pathologist who has heard you. Team Gleason says banking remains possible even after voice changes begin or the voice is substantially lost, and cloning from existing audio is a separate route for people who cannot record new material. Ask before you assume.
What is the difference between voice banking and message banking?
Voice banking builds a synthetic voice that can say anything you type later. Message banking saves specific phrases in your real voice, played back exactly as recorded. Most guidance recommends doing both, and Boston Children’s Hospital describes double dipping, where one set of banked messages is reused to build a synthetic voice, so the effort is not duplicated.
Does insurance or a charity pay for this?
Team Gleason says anyone living in the USA with an official ALS diagnosis is eligible for its funding, and that it covers the cost of the voice banking software chosen, working with the voice banking companies the ALS Association lists. Apply before you pay anything, because Team Gleason states it cannot reimburse expenses made before approval.