Written from 22 named sources · Oct 16 · first result A research companion to VO Infusion Episode 3, "From Fear to First Prompt." Researched with Pig Knuckle and reviewed by Deb Munro and Keith Norton, with sources for every claim. Key findings Voice banking records prompted phrases so software can build a synthetic voice that speaks whatever you type. Message banking saves exact recordings in your natural voice for playback [1][3]. Clinicians say to start while your voice is strong, because what goes in shapes what comes out [2][3][5]. Cost runs from free (Apple Personal Voice, and ElevenLabs for eligible patients) to about US$1,000 for an Acapela flat fee [1][3][13]. Published prices for the same vendor don't always agree. Recording loads range from about 10 phrases on newer iOS versions to 400 or more for ModelTalker, depending on the source [1][3]. Traditional voice banking generally gives you a voice file that works offline. An AI clone lives on the company's server, and as one ALS clinic speech-language pathologist put it, "you don't own a file" [3]. A 2025 licensing analysis reports that ElevenLabs lets paid users keep rights to generated audio but takes a perpetual, irrevocable license over submitted voice data [14]. A career-insurance clone needs an archive you control, written ownership of the trained model, API keys in your name and a named successor. None of the accessibility programs I reviewed package all of that for working performers [1][3][13], so treat it as a separate project with its own contract. US disability coverage built for entertainers exists, including a published voice actor example that replaces 65% of income [15]. The policy's definition of disability decides whether it pays, and public and union plans tend to use much stricter tests [10][16][18]. As of October 2026, the NO FAKES Act (S.4591) cleared the Senate Judiciary Committee in June. It stalled on September 30, when Sen. Ted Cruz objected to passing it by unanimous consent, so it is still a bill [7][21]. Tennessee's ELVIS Act is state law today, and Canada relies on a provincial patchwork [7][9]. Voice banking, message banking and what they cost I stepped away from TTS work to protect my voice. I've also watched friends lose their voices to ALS and end up typing through the same synthetic voice... different people, one stock sound. So this piece starts with the people who need these tools most, then asks what it would take for the rest of us to own a copy of our own instrument. Voice banking means recording a set of phrases a service gives you, so software can build a synthetic version of your voice. That voice then speaks anything you type through text-to-speech (TTS) [1][3]. It usually runs on a speech-generating device, the clinical term for a communication device that talks, or in an AAC (augmentative and alternative communication) app [4]. Message banking is simpler. You record greetings, names, jokes and your laugh, and they play back exactly as you said them [1]. The Sunnybrook clinicians on ALS Canada's webinar note that message banking has no minimum, no speech-quality cutoff, and works in any language you can record [3]. Voice cloning is the newest branch. It uses AI and almost any clean recording of you, including old voicemails and videos [3][13]. Timing matters, because a stronger voice at recording time usually produces a better result [2][5]. If speech has already changed a lot, a sibling who sounds like you can bank and donate a voice, and old recordings can feed a clone [3]. A speech-language pathologist (SLP) should confirm that the voice will work with your communication device [5]. The table below shows the main options as of October 2026. Where Preserve Your Voice [1] and the ALS Canada webinar [3] list different prices or recording loads, I've included both, so confirm with the vendor before you pay. Option Recording needed Cost (USD) What you end up holding Message banking No minimum [3] Free to a few hundred dollars [1] Your own recordings, usable offline [3] Apple Personal Voice 150 phrases, about 10 minimum on iOS 26 [1][3] Free on supported devices [1] A voice across your Apple devices [3] Voice Keeper 5 sentences [3] or 50+ phrases [1] $5 a month [3] or $299 [1] Voice file, offline use [3] Acapela My Own Voice 50 phrases or longer sentences [1][3] $99 a year or about $1,000 flat [1][3] Voice file, offline use [3] SpeakUnique 125 to 150+ sentences [1][3] $190 to $270 [1] or $400 to $686 [3] Voice file, offline use [3] ModelTalker 300 to 400+ sentences, 1 to 2 hours [1][3] $100 [1][3] Voice file, offline use [3] ElevenLabs Impact 10 to 30 seconds instant, 10+ minutes professional [3][13] Free Pro license, valued at $1,200 a year [13] Cloud clone, internet required [3][13] Peter Jonkers, writing about his own ALS experience, found that AI clones sound nearly human while older systems sound more robotic [2]. Medical insurance generally doesn't pay for voice or message banking [1]. In the US, Team Gleason sponsors the cost for people with ALS in North America [1], and the VA should cover software for eligible veterans [5]. Bridging Voice helps US applicants get the free ElevenLabs license with a letter from a doctor or SLP confirming the diagnosis [13]. Canadians apply through the Scott Morgan Foundation option for that same program [3]. Ontario runs a government-funded equipment pool that evaluates new communication technology [3], so ask your ALS clinic what your province covers. Who owns a banked or cloned voice Three different things get called "your voice" in these deals. The first is the source recordings. The second is the trained model, the file of numbers called model weights that actually is the clone. The third is the right to sell what that model produces. Holding your recordings gives you no automatic claim to the weights, and holding the weights doesn't by itself give you a commercial license. The ALS Canada clinicians laid out the practical split plainly. With message banking, you record on your own device, so "you kind of own it already." Traditional voice banking gives you a sound file you download and use offline. With a clone, you upload recordings, the company's server builds the model, and every new sentence travels to that server and back [3]. Preserve Your Voice confirms that an ElevenLabs Impact clone is stored in the ElevenLabs database [1]. Bridging Voice tells patients the voice is used only to help them communicate and isn't shared without permission [13]. Seriously: read the terms before you upload. LicenseOrg's analysis reports that paid users keep rights to the audio they generate and that ElevenLabs promises not to commercialize a voice on a standalone basis without permission. It also reports that the terms grant the company a perpetual, irrevocable, non-exclusive, royalty-free, worldwide, sub-licensable license over the voice model for service improvement, and that users pushed back in early 2025 [14]. You can opt out of training through the "Data use" menu in account settings [14]. That analysis reflects early 2025 terms, so pull the current version from ElevenLabs itself. Apple Personal Voice is documented as a free communication feature that works with FaceTime, Live Speech and a growing list of AAC apps [1][5]. None of the clinical sources I reviewed describe an export option, a portable model file or a commercial license, so check Apple's own support pages before you treat it as anything more than a personal backup. For a patient, that limit may not matter. For a performer, it's a wall. Legacy access varies too. ALS Canada's presenters said some banking platforms delete voices after a number of years, while others offer a legacy setting so family can reach the voice later. For ElevenLabs, they weren't sure about a legacy feature. The control they could point to was disabling your API key [3]. They also guessed, without certainty, that US-based services fall under US privacy law, which Canadian users should weigh [3]. Before you record anything, get written answers on four questions. Who owns the model? Can you export it or switch providers? Who can manage it if you can't? Does your audio train anything else? What a performer-owned clone would take On VO Radio, a 23-year-old caller asked how she was supposed to live in her own society if veterans told her to run from AI. Deb's reaction was "we're being dinosaurs here." Pair that with the idea from the episode of a clone you own, API and all, that could keep a working voice actor earning after a stroke or illness. Start with an archive you control, built while you're healthy. Preserve Your Voice recommends WAV files at 44.1 kHz and 16-bit for banked messages [1], and that's a sensible floor for performers too. Add conversational reads, narration and the character voices you've actually booked, and keep one copy local and one in a separate cloud account in your name. Do you know where your session WAVs from 2019 are right now? I'd need a minute on that one myself... The contract should name you as owner of both the recordings and the model weights, and limit the provider to technical services. It should bar training other models on your audio and selling access to anyone else. Commercial permission works best one project at a time, with the client, medium, territory, term and your approval rights spelled out. Any "perpetual and irrevocable" grant turns your insurance into the vendor's inventory. An API, or application programming interface, is the doorway that lets other software ask a service to generate audio in your voice. API keys are the passwords to that doorway. ALS Canada's team described unique API keys, watermarking on professional clones and the option to switch a key off [3]. Watermarking is a tag that marks audio as synthetic. Production volume matters as well, because Pro accounts carry character limits per message and per month [3][13]. The diagram below shows how those pieces connect when control stays with the performer. The successor box is the part most of us skip, and it's the one that decides whether the clone outlives your ability to run it. The skeptic's case deserves a fair hearing. A clone you own could still undercut your live bookings, and a platform's training license might give away more than the clone protects [14]. Clients and unions may refuse synthetic delivery altogether. As of October 2026, I found no public data showing performers earning real income from a personal clone after losing their voice. Tim Friedlander of NAVA walked away from a production that went almost all AI. Investment in human production also continues: Paramount signed a minimum 10-year, 285,000-square-foot lease at 1888 Studios in Bayonne, New Jersey, a campus Variety reports should be operational by the fourth quarter of 2028 [19][20]. My answer to the skeptics is that a clone only works as insurance if the contract keeps the off switch in your hand. Disability insurance for voice professionals Most working voice actors probably have no disability insurance. That's my read from the episode, not a survey result. Disability insurance replaces part of your income when illness or injury stops you from working, usually 60 to 65% [16]. The definition of disability decides whether a claim pays, and the table below compares the versions you'll run into. Definition Pays when Where you'll see it True own-occupation You can't do the material duties of your occupation, even if you earn elsewhere [18] Individual policies [18] Modified own-occupation Same test, but only while you're not working another job [18] Some individual policies [18] 24-month switch Own-occupation for about two years, then any-occupation [10][18] Group plans and some Canadian base policies [10][18] Any-occupation You can't do any job your training fits you for [18] US SSDI under 20 CFR 404.1505 [18] CPP Disability A severe and prolonged disability leaves you unable to work at any occupation [10] Canada's public plan, 2026 maximum $1,741.20 a month [10] Two more clauses matter to us. Residual disability pays a share when you lose part of your income. One RBC policy, for example, triggers it at a 20% loss of earnings [10]. That covers the case where a vocal-fold condition ends your VO work but not your on-camera work. Presumptive disability treats loss of speech as total disability even if you keep working [18], so ask for that clause by name. Specialty markets also write voice-specific riders for broadcasters and vocalists [17]. Products do exist. Petersen International Underwriters publishes an entertainment-industry plan with a voice actor example, offering a plan that would replace 65% of his $125,000 income [15]. That's marketing copy, so ask for the actual policy wording. A 2018 Edge Studio article reported that SAG-AFTRA's disability pensions appeared to cover only total disability, defined as unsuitability for "any gainful employment" [16]. Confirm that against the current plan documents. In Canada, RBC's public pages show the vocabulary you'll meet, including a Regular Occupation Extension rider that keeps the own-occupation test past 24 months [10]. They also show exclusions. RBC's simplified policy excludes mental health and subjective conditions [10], so ask how throat, lung and neurological conditions are treated. ACTRA members should ask ACTRA about member benefits. Nothing here is financial advice, and a licensed broker who understands self-employed creative income should quote your actual coverage. Where the NO FAKES Act, the ELVIS Act and Canadian law stand The timeline below covers the dates that matter for performers. One discrepancy belongs on the record: most sources date the committee vote June 18, 2026 [7][12][22], while Byte Back reports June 22 [9]. The NO FAKES Act (Nurture Originals, Foster Art, and Keep Entertainment Safe) would create a federal right to authorize digital replicas of your voice and likeness. It would also add notice-and-takedown rules for platforms, with damages of up to $750,000 per work for platforms that make no good-faith effort to comply [7]. Statutory damages for violators would run from $5,000 to $25,000 per replica [9]. On September 30, 2026, Sen. Marsha Blackburn asked the Senate to pass it by unanimous consent. Sen. Ted Cruz objected that the parody exception leaves Big Tech to decide what counts, and he said his staff would work with the sponsors [21]. For families, Deadline reports that the proposed right would survive death, could be licensed or transferred by heirs and would end no later than 70 years after death [22]. Paul Hastings notes the bill restricts lifetime assignment and leans on consent-based licensing, so voice rights may not behave like copyrights [11]. Tim Friedlander, NAVA's president, said NAVA took part in over 100 meetings supporting the bill [22]. Tennessee's ELVIS Act sits at Tenn. Code Ann. §§ 47-25-1101 to -1108 [7] and was signed March 21, 2024 [11]. It creates a private right of action against tools whose primary purpose is producing an identifiable person's voice or likeness without consent [9]. Because it predates the January 2, 2025 cutoff, analysts expect it to survive federal preemption [9]. Eight states now have digital-replica laws [9], and California updated its impersonation law to cover digital replicas on September 30, 2026 [21]. Canada is where my sources run thin, so treat this paragraph as a map for a lawyer conversation. As of October 2026, I found no federal digital-replica bill. Quebec's Civil Code protects personality rights in articles 3, 35 and 36. Ontario courts recognize a tort of appropriation of personality, and several provinces have statutory privacy torts. ACTRA members should ask ACTRA for current AI contract language, and anyone facing a specific deal should ask a Canadian lawyer. What this means for your voice business The episode's line for this part was "build a philosophy before you buy a subscription." Each of these steps fits inside a month. Write a one-page personal AI policy that covers training, commercial use, synthetic pickups and use after your death. Share it with your agent and one person who could speak for you. Gather the clean sessions, scripts and finished work you have rights to keep, saved as WAV at 44.1 kHz and 16-bit or better [1]. Store one copy locally and one in a separate account in your name, along with a log of dates, microphones and client permissions. Send two providers the same written questions on model ownership, export, API keys, deletion and estate control, and opt out of training wherever the platform allows it [3][14]. Ask NAVA or your union for current synthetic-voice contract language, and have a lawyer read any AI deal before you sign. Call a licensed broker about true own-occupation, residual and presumptive-disability coverage, plus any voice-specific rider [17][18]. Canadians should compare quotes against CPP's "severe and prolonged" test and ask ACTRA about member benefits [10]. Contact your senators about S.4591 and your House member about H.R.8915, since the bill stalled on September 30 [21]. Read Tennessee's ELVIS Act so you know what state protection looks like [7], and Canadians can carry the same message to their MP and through ACTRA. Go further ALS Canada Voice Preservation webinar [3]: Sunnybrook clinicians demo every major method, including the ownership question from the audience. It's the best place to hear the differences. Preserve Your Voice [1]: plain steps, vendor prices and recording settings for banking and message banking. It's good to share with families. Bridging Voice's ElevenLabs page [13]: eligibility rules and audio requirements for the free Pro license. Check it before applying. LicenseOrg's ElevenLabs licensing guide [14]: a readable breakdown of output rights versus the voice-model license. Verify it against the current terms. Seaworthy's own-occupation explainer [18]: the clearest walk-through of disability definitions I found. Bring it to your broker call. Starchild's account of the Senate objection [21]: the September 30 floor fight and what states are doing meanwhile. It's the fastest catch-up on current status. Cardozo Arts and Entertainment Law Journal [7] and Deadline [22]: the committee timeline, damages and the post-mortem provision. Read both if you want the full legislative picture. Pig Knuckle (pigknuckle.ai) is a research orchestration platform from Alchemy Agentic, founded by Paul Langtry and Keith Norton. It breaks a research question into steps, sends each step to the AI model best suited to it, and builds its citation list from the documents it actually read, so every claim can be traced to a source. VO Infusion is a podcast about artificial intelligence and the voice acting industry, co-hosted by voice actor and coach Deb Munro and voice actor Keith Norton. Episodes cover voice cloning and licensing, legal protections for performers, prompt engineering for voice actors, AI tools for the business side of voiceover and why human performance still matters. Listen wherever you get podcasts or on YouTube at @VOInfusion, and find more at voinfusion.com. Sources [1] preserveyourvoice.com — https://preserveyourvoice.com [2] #als #voicebanking #accessibility #ai #patientexperience #digitalhealth #elevenlabs #speakunique #acapela #applepersonalvoice #mindspeaker #scottmorganfoundation Peter Jonkers — https://www.linkedin.com/posts/peterjonkers_als-voicebanking-accessibility-activity-7363813719044227072-_tVu [3] ALS Canada Webinar: Voice Preservation — https://www.youtube.com/watch?v=aCKFXf_6ygc [4] Voice Banking: Preserving Your Voice Through Assistive Technology Ability Central — https://abilitycentral.org/article/voice-banking-preserving-your-voice-through-assistive-technology [5] Voice Banking - Your ALS Guide — https://www.youralsguide.com/voice-banking.html [7] Borrowed Voices: What the NO FAKES Act Means for AI Disclosure Risk - Cardozo AELJ — https://cardozoaelj.com/2026/10/12/borrowed-voices-what-the-no-fakes-act-means-for-ai-disclosure-risk [9] A Federal Shift in AI and the Right of Publicity? NO FAKES Act Advances in Congress Byte Back — https://www.bytebacklaw.com/2026/08/a-federal-shift-in-ai-and-the-right-of-publicity-no-fakes-act-advances-in-congress [10] Disability Insurance in Canada - Get a Quote RBC Insurance — https://www.rbcinsurance.com/en-ca/health-insurance/disability-insurance [11] The Next Music Asset: How the NO FAKES Act Could Reshape Music Licensing, Investment and Commercialization in the AI Era Paul Hastings LLP — https://www.paulhastings.com/insights/client-alerts/the-next-music-asset-how-the-no-fakes-act-could-reshape-music-licensing-investment-and-commercialization-in-the-ai-era [12] Senate Committee Advances Bill to Protect Name, Image, Likeness and Voice Against Unauthorized AI Use Insights Holland & Knight — https://www.hklaw.com/en/insights/publications/2026/06/senate-judiciary-committee-advances-legislation-to-protect-name [13] ElevenLabs - Bridging Voice — https://bridgingvoice.org/elevenlabs/ [14] ElevenLabs Licensing Guide: What You Can (and Can't) Do With AI Voices — https://www.licenseorg.com/blog/elevenlabs-licensing-guide-ai-voices [15] Personal Disability Insurance Plan - Entertainment Industry — https://www.piu.org/personal-disability-insurance/actors-actresses [16] Does a voice actor need insurance? Yes, and no. And Yes. — https://edgestudio.com/does-a-voice-actor-need-insurance-yes-and-no-and-yes/ [17] Disability Insurance for the Entertainment Industry — https://www.diversifiedquotes.com/disability-insurance-for-the-entertainment-industry/ [18] Own-Occupation Disability Insurance Explained — https://seaworthy.io/education/own-occupation-disability-insurance/ [19] Paramount Signs 10-Year Lease With 1888 Studios For New Jersey Production Campus — https://variety.com/2025/film/news/paramount-signs-10-year-lease-1888-studios-1236564020/ [20] Paramount To Anchor New 1888 Studios In Bayonne, New Jersey — https://deadline.com/2025/10/paramount-1888-studios-bayonne-new-jersey-1236600571/ [21] The NO FAKES Act Stalls in the Senate: Where Voice Rights Stand — https://starchild.music/blog/no-fakes-act-blocked-senate-voice-rights [22] No Fakes Act Clears Senate Judiciary Committee — https://deadline.com/2026/06/no-fakes-act-senate-judiciary-committee-1236959147/ Elevenlabs vs Voice Keeper vs Modeltalker for voice banking — https://www.reddit.com/r/ALS/comments/1i44o4p/elevenlabs_vs_voice_keeper_vs_modeltalker_for What AI recommends for voice preservation and legacy archive services — https://essentras.com/ai-check/category/voice-preservation-and-legacy-archive-services