Laryngectomy Voice: AI Cloning & Voice Changer Options After Surgery
Losing your voice to laryngectomy is not a small thing. For many people, it follows a cancer diagnosis — a period of fear, treatment, and upheaval — and the surgery itself removes the organ that produced the sound most people never think about until it’s gone. The voice you used to laugh, argue, comfort, and simply talk your way through daily life is changed, sometimes irreversibly, in a matter of hours.
This guide is about what technology can offer in that situation — honestly, without overselling. AI voice cloning and voice changer software have advanced enough to be genuinely useful for some laryngectomy patients, particularly as a complement to traditional alaryngeal speech methods. But they are one option among several, and they work best alongside professional rehabilitation, not instead of it.
TL;DR
- Laryngectomy removes the larynx; three established methods replace voice: electrolarynx, esophageal speech, tracheoesophageal voice prosthesis (TEP).
- AI voice banking — recording your voice before surgery — creates a personal voice asset that AI tools can draw from afterward.
- AI voice changers can process electrolarynx or TEP audio in real time, making it sound less robotic.
- Results are meaningful, not magical: improvement is real, restoration is not the right word.
- Work with a speech-language pathologist (SLP). Technology supports rehabilitation; it does not replace it.
- Organizations: WebWhispers, International Association of Laryngectomees, ASHA.
What Happens to the Voice After Laryngectomy
The larynx — the voice box — contains the vocal folds that vibrate to create sound. During a total laryngectomy, the entire larynx is removed, the trachea is rerouted to a permanent opening in the front of the neck (the stoma), and the connection between the lungs and the mouth is severed. Breathing and voicing no longer share the same path.
Without the vocal folds, the voice as it was does not exist. What replaces it depends on anatomy, cancer staging, reconstruction choices, and personal preference — but the journey begins with rehabilitation and often continues for years.
The emotional weight is significant. Research on laryngectomy quality of life consistently documents grief, social withdrawal, and identity disruption alongside the practical communication challenges. Technology is not the answer to those dimensions of loss, but it can reduce some of the daily friction.
The Three Established Alaryngeal Speech Methods
Before discussing AI tools, understanding what most laryngectomees work with in rehabilitation gives important context.
| Method | How It Works | Pros | Cons |
|---|---|---|---|
| Electrolarynx | Handheld device vibrates neck/cheek; the mouth shapes the sound | Easy to learn, reliable | Robotic buzz, requires one hand free, proximity to neck |
| Esophageal Speech | Air is trapped and released through the esophagus to create vibration | No device needed, hands-free | Long learning curve, low volume, fatiguing |
| TEP (Tracheoesophageal Voice Prosthesis) | Surgically placed valve; lung air creates voice through prosthesis | Best volume and naturalness, hands-free possible | Requires surgery, prosthesis maintenance, fitting appointments |
None of these methods is inherently superior. The right choice depends on factors your surgical team and SLP will evaluate: radiation history, anatomy, age, occupation, personal goals. Many people use more than one method in different situations.
The electrolarynx is often the first method introduced post-surgery because of how quickly it can restore basic communication. Its characteristic robotic sound is also the starting point where AI voice processing becomes relevant.
Voice Banking: The Case for Recording Before Surgery
If there is one message in this guide worth sharing before a scheduled laryngectomy, it is this: record your voice now, before surgery, if at all possible.
Voice banking is the process of recording a large set of speech samples — sentences, words, conversational phrases — so that AI models can learn the characteristics of your voice. The larger and more varied the recordings, the better the resulting AI model can capture your natural timbre, pacing, and intonation.
Projects like Project Revoice demonstrated with ALS patients — including co-founder Pat Quinn — that banking a voice before it is lost creates a personal voice asset that remains usable afterward. The principle applies directly to laryngectomy: a voice banked pre-surgery can be used as the target voice in AI cloning software, giving the processed output a personalized quality rather than a generic one.
What you need for voice banking:
- A quiet room and a decent microphone (a USB headset is sufficient)
- At minimum 15 minutes of varied speech — sentences, questions, counting, reading aloud
- More is better: 1-2 hours of recordings substantially improves AI model quality
- Variation: different emotional tones, speeds, and content types help
If surgery is weeks away, this is achievable. If the timeline is very short, even a few hours of recorded material is worth having. Contact your SLP team — many hospitals now have voice banking protocols, and some have partnerships with voice banking services.
What AI Voice Changers Actually Do for Laryngectomy Patients
After surgery, AI voice changer software addresses a specific gap: the electrolarynx produces sound that is clearly artificial — its monotonal buzz is a marker that many users find socially limiting. AI voice conversion takes that input and processes it in real time, applying a learned voice model to make the output sound warmer, more varied in pitch, and more human in character.
Here is what this looks like in practice:
- The user speaks with an electrolarynx (or through TEP-produced speech) into a microphone near their mouth or throat.
- The AI voice changer software captures that audio, processes it through a voice conversion model in real time.
- The processed output — less robotic, closer to the target voice profile — goes to a virtual microphone.
- Any call app, video conferencing tool, or communication software receives the virtual microphone’s output.
This is not synthesis from text. It is real-time acoustic transformation of the actual speech signal. Articulation, rhythm, and the words themselves come from the user; the software changes the tonal quality and timbre.
The honest assessment: the improvement is real and often significant, but it is not restoration. Users consistently report that processed electrolarynx speech is easier for listeners to understand and less emotionally marked by its robotic quality. They do not report sounding identical to their pre-surgery voice. Expectations matter enormously here.
AI Voice Cloning: Using Pre-Surgery Samples
If voice banking was done before surgery, AI voice cloning takes the concept further. Instead of applying a generic voice conversion model, the software is trained on — or fine-tuned with — the user’s own pre-surgery recordings. The result is a personalized voice model rather than a generalized one.
VoxBooster supports this workflow: upload 15 or more minutes of pre-surgery audio samples, train a personal voice model, and use it for real-time conversion. The output reflects the acoustic characteristics of your specific voice — its natural warmth, resonance, and character — rather than a neutral baseline. For people who banked recordings before surgery, this is the closest existing technology can come to personal voice continuity.
This is not available to everyone. Many patients receive their diagnosis with limited lead time. The voice banking window may not be long enough, or it may not have been offered. In those cases, a general-purpose voice model can still improve the naturalness of electrolarynx output — the gain is just less personalized.
Practical Setup: Getting AI Voice Processing Running
For Windows 10/11 users, the setup for real-time AI voice conversion with an electrolarynx or TEP is straightforward:
Hardware you need:
- A small microphone positioned near the neck/mouth (a lapel mic or a close-mounted condenser works well)
- Standard audio interface or USB microphone input
- Windows 10 or 11 PC — does not need to be high-end; modest CPU is sufficient for most AI voice processing
Software setup with VoxBooster:
- Install VoxBooster — no kernel driver is installed, which keeps CPU load lower and avoids compatibility issues with older PCs
- Set your physical microphone as the input
- Select your voice model (pre-trained on your banked samples, or a general model)
- Set VoxBooster’s virtual microphone as the input in your call app
- Speak — conversion happens with low enough latency for natural conversation
A note on latency: Real-time voice conversion introduces a small delay, typically 100-300ms depending on hardware and model complexity. This is perceptible but manageable for most conversations. For in-person communication where the physical voice is also present, the delay is more noticeable; the workflow is optimized for phone calls, video calls, and online communication.
VoxBooster also includes Whisper-based transcription as a fallback option — useful for situations where voice clarity is uncertain, allowing text-based communication alongside or instead of voice conversion.
Comparison: Alaryngeal Speech Methods and AI Enhancement
| Method | Sound Quality | Learning Effort | Hands-Free | AI Enhancement Possible |
|---|---|---|---|---|
| Electrolarynx | Robotic but intelligible | Low — fast to learn | No (one hand used) | Yes — significant improvement |
| Esophageal Speech | More natural but low volume | High — months of practice | Yes | Possible but less common |
| TEP | Best overall quality | Moderate — valve management | Mostly yes | Yes — subtle refinement |
| AI Cloning (banked voice) | Personalized, warmer | Setup only | Yes (via virtual mic) | N/A — is the enhancement layer |
AI voice processing is most impactful as an enhancement on top of electrolarynx, which is the most commonly used method. It is also compatible with TEP speech for users who want further refinement on calls.
The Role of Your Speech-Language Pathologist
This section exists because it is easy to read about technology and conclude that the path forward is software. It is not — the path forward is rehabilitation, and software is one tool within that.
A speech-language pathologist is the professional who:
- Evaluates which alaryngeal speech method is appropriate for your anatomy and situation
- Teaches proper technique (improper electrolarynx use reduces intelligibility and can cause discomfort)
- Adjusts the rehabilitation plan as you progress
- Coordinates with your surgical team on prosthesis management if using TEP
- Can advise on whether AI voice software is appropriate and how to integrate it
- Addresses the emotional and social dimensions of communication change
The American Speech-Language-Hearing Association (ASHA) maintains a directory of certified SLPs and resources for patients. If you are outside the US, national SLP associations exist in most countries with links to local practitioners.
AI voice changers and cloning tools do not require a prescription and do not replace a SLP — they are software anyone can try. But the decision about whether and how to use them benefits from professional guidance.
Community and Support Resources
Laryngectomy rehabilitation is not a solo journey. Support communities provide practical knowledge — which microphone positions work best with an electrolarynx, which call apps work most reliably with virtual microphone setups, how other people have adapted AI voice tools to their workflow — that no guide can fully anticipate.
WebWhispers is the largest online support network for laryngectomees in English. Its forums and email lists have been running for decades and contain a remarkable depth of practical experience.
International Association of Laryngectomees (IAL) connects local Lost Cord and New Voice clubs around the world, with chapters in dozens of countries. Many chapters run in-person support meetings.
ASHA and national equivalents provide clinician directories and patient-facing educational materials.
If you are researching AI voice tools on behalf of a family member or patient, these communities are the first place to ask about what is actually working for people in practice.
Honest Limitations
This guide would not be complete without stating clearly what AI voice technology cannot do:
- It cannot fully restore the pre-surgery voice. Voice cloning from banked samples produces a personalized approximation; it is not the original voice.
- It does not work well in noisy environments. Background noise degrades conversion quality significantly.
- Latency is real. Real-time processing introduces delay that some users find disorienting.
- It requires a working Windows PC and a reasonable level of technical comfort for initial setup.
- Emotional and social adjustment to voice loss is not solved by software. That work is human work — therapy, support groups, time.
The goal of AI voice tools for laryngectomy is to reduce daily friction in communication, particularly in phone and online contexts where the physical context of face-to-face conversation is absent. That is a real and meaningful goal. It is also a modest one.
Getting Started
If you or someone you care for is preparing for or recovering from laryngectomy:
- Before surgery if possible: Record at least 15-60 minutes of natural, varied speech for voice banking. Contact your SLP team about voice banking protocols — many hospitals now have them.
- After surgery: Work with your SLP to establish the primary alaryngeal speech method appropriate for your situation.
- When stable in rehabilitation: Explore AI voice software as a complement — particularly for phone and video call contexts where the robotic quality of an electrolarynx is most limiting.
- Connect with community: WebWhispers and your local IAL chapter have users who have navigated this process and can share practical experience.
VoxBooster offers a 3-day free trial (no credit card required) for anyone who wants to test real-time AI voice conversion before committing. At $6.99/month, the cost is low enough to explore seriously. It runs on Windows 10 and 11, requires no kernel driver installation, and supports personal voice model training from your own recordings.
Voice loss is profound. The tools available to help are imperfect. But the combination of good rehabilitation, professional support, and carefully chosen technology has helped many people rebuild a communication life that works — different from before, but meaningful.
This post is informational and does not constitute medical advice. Always consult qualified medical and speech-language pathology professionals for decisions about laryngectomy rehabilitation.