Terms of Service

Voice Buddy is operated by Workflow Corporation, which trades as Radio Workflow HQ and has its place of business at 210 Emerson Pl, Suite 300, Davenport, IA 52801, USA. In this document “we”, “us” and “our” mean Workflow Corporation; “you” means the person or organisation using Voice Buddy. You can reach us at support@workflowcorp.com or +1 (563) 275-6409; support is available at any hour, and the office is open Monday to Friday, 8:00 AM to 6:00 PM Central, closed at weekends.

These terms apply to the Voice Buddy website, the console, and the REST API. Two further documents form part of this agreement and are incorporated by reference: the Acceptable Use Policy and the Voice Cloning and Biometric Data Notice. The Privacy Policy explains what we do with data and is not part of the contract but describes how we perform it.

You accept these terms by creating an account, by joining a workspace you were invited to, or by using the API with a key issued to you. If you are accepting on behalf of an organisation, you confirm you have authority to bind it.

You must be at least 18 years old to hold an account. Not 13, and not “the age of digital consent where you live” — 18 everywhere. This service takes payment by card and it processes voice recordings, which are biometric identifiers; neither a subscription nor a consent to have a voiceprint made is something a child can give. We do not knowingly provide the service to anyone younger, and if we learn that an account holder is under 18 we will close the account and delete what it holds. This is a rule about who may hold an account, not about whose voice may be cloned: a child’s voice may be used only on the consent of a parent or guardian, which the Voice Cloning and Biometric Data Notice sets out.

All content in Voice Buddy belongs to a workspace, not to an individual. A workspace has members, and each member holds one of four roles:

Owner

Full control, including deleting the workspace and everything in it.

Admin

Manages members, voices, billing and API keys. Can start a professional voice training run, which spends our compute rather than your allowance.

Member

Can generate audio, which spends the workspace’s allowance.

Viewer

Read-only. Sees the workspace’s voices and history but cannot generate.

Everything a member creates is visible to the workspace. Generation history, uploaded audio, transcripts and scripts are workspace-scoped, so anyone with access to the workspace can see them. Do not put anything in a shared workspace that you would not show its other members.

You are responsible for keeping your password and API keys secret and for everything done through them. API keys are hashed at rest, so we cannot show you a key again after it is created — if one is lost or exposed, revoke it and issue a new one.

Input means everything you put into Voice Buddy: scripts and text, audio and video you upload, voice samples, music beds, and the settings you choose. Output means the audio, transcripts, translations and scripts the service produces for you.

You keep whatever rights you already had in your Input. As between you and us, and subject to the plan licence below and to your compliance with these terms, you own the Output produced from it.

The licence you grant us

You grant us a worldwide, non-exclusive, royalty-free licence to host, store, copy, transmit and process your Input and Output for the purposes of operating the service for you: rendering it on our speech engine, storing it in our object storage, sending the parts that need a third-party model to the providers described in the Privacy Policy, and showing it back to your workspace. This licence exists so we can run the product; it ends for a given item when that item is deleted, subject to the backup window described in the Privacy Policy.

Training on your renders — read this one

Voice Buddy has a professional voice cloningfeature. When an admin of your workspace starts a training run for one of your voices, the service builds a dataset from that workspace’s own finished renders of that same voice — the audio together with the script that produced it — and a GPU job fine-tunes an adapter on it. That adapter is attached to your voice and is used to render your voice.

Three things are true about this today and we would rather say them plainly than bury them:

• The dataset is scoped to the one workspace and the one voice. Your renders are not pooled with other customers’ renders, and they are not used to train the shared speech engine or any voice you do not own.
• There is no separate consent step and no per-render opt-out. Every successful render of a voice, kept in history, is eligible material for a future training run on that voice. Deleting a generation from history removes it from future datasets; it does not unlearn an adapter that was already trained.
• The training job runs on our infrastructure and reads your stored audio from our storage. It is not sent to a third-party model provider.

We do not sell your Input or Output, and we do not use it to train models offered to anyone else.

You warrant that you have all rights, consents and permissions needed for everything you upload and everything you generate. In particular:

• For any voice you clone, you either are the speaker or hold the speaker’s express, informed, documented consent to clone and synthesise their voice for the uses you intend. The Voice Cloning and Biometric Data Notice sets out what that consent has to cover.
• For any audio or video you upload to re-voice, transcribe, dub or mix, you hold the rights to the recording, the underlying work and the performances in it.
• For any music bed you upload, and for any sound effect you take from the Freesound library through the search surface, you comply with its licence — including attribution where the licence requires it. We record the licence and author we were given; complying with it is yours to do.

This is the single most important promise you make to us, because it is the one we cannot verify. You agree to indemnify us against claims arising from a breach of it — see Indemnity below.

The Acceptable Use Policy lists what you may not do with the service. Breaching it is a breach of these terms. Impersonating a real person without their consent, using synthetic speech to defraud or harass, and generating deceptive political content are among the things that will get an account suspended immediately rather than warned.

One case is worth naming here rather than leaving to the policy, because it is regulated rather than merely prohibited: political and election advertising. Where synthetic or AI-generated voice is used in it, disclosure rules apply — they differ by where the ad runs and they are not ours to interpret for you. You are responsible for complying with every rule that applies to your ad, and for disclosing that the voice is synthetic. We do not add that disclosure to your audio and cannot tell from a render whether one is required.

Each plan carries a monthly character allowance and a cap on the number of custom voices. Both are enforced: a render that would take you past the allowance is refused, and so is a voice past the cap. The current figures are on the pricing page and in the console; the Free plan is 10,000 characters and 3 custom voices.

• Allowances do not roll over. The counter resets to zero at the start of each billing period. Unused characters are not credited, refunded or carried forward.
• Plans also list a concurrency figure. We do not enforce it today. Nothing counts your in-flight renders and nothing refuses a request for being simultaneous, so that number is neither a cap we apply to you nor a throughput we promise you — do not size a worker pool from it. If we start enforcing it, that is a material change to a paid plan and is notified as one.
• Text is metered by characters submitted. Speech-to-speech, transcription and dubbing are metered by the audio they process. A render that fails on our side is not charged against your allowance.
• The Free plan is for evaluation and non-commercial use only. Audio generated on the Free plan may not be used in anything you sell, advertise with, or monetise. Commercial use requires a paid plan.
• Every request made with an API key is metered against the workspace that owns the key, whoever made it.
• A few features carry a separate per-period ceiling of their own, because they spend a third party’s money rather than our GPUs. Generated backgrounds and storyboards, designing a voice from a description, live conversation with a voice agent, and professional voice cloning each have a limit on how many you may start in a billing period. None of them consumes your character allowance and none of them moves the character counter, so that is not the number to watch for them: the console states what is left beside the control that spends it, and the API publishes it before you make the call that spends it — on the subscription read, on the video-template read, and, for cloning, on the training read for the voice. These ceilings scale with the plan and reset on the same day the character allowance does. They are fair use, not a purchased quantity — we may change them, and we will tell you if we lower one on a paid plan.
• Live conversation is the one of those to read twice, because running out of it does not look like running out of the others. It is counted in minutesrather than in presses; it spends continuously for as long as somebody is talking; and where you have pointed one of your own telephone numbers at an agent, it is the only allowance in this product that a member of the public can spend without anybody at your end doing anything. When it runs out, a conversation already in progress is ended and a new one is refused — which to a caller is a line that does not work rather than a message about an allowance, so it is worth watching the way you would watch a prepaid balance and not the way you watch the character counter. One conversation is separately capped at 30 minutes however much the two parties still had to say, and that stop is enforced at both ends, so it happens whether or not your browser is still there.

The commercial licence, on a paid plan

On any paid plan we grant you a worldwide, non-exclusive, royalty-free, perpetual licence to use the Output commercially: to reproduce, edit, combine, distribute, broadcast, perform, publish and monetise it, in any medium and for any client, with no per-use fee and no obligation to credit us or to say which tool made it. Non-exclusive means we grant the same rights to every other customer — it does not mean anyone else gets your audio, which we do not give to anyone. You may sublicense the Output as part of a finished production — an advertisement delivered to a client, an episode, a course, a game — and that sublicence survives cancellation, so work you have already shipped does not become unlicensed when you stop paying.

Three limits, and they are the only ones:

• The Free plan carries no commercial licence at all. The plan is listed as non-commercial where you sign up for it, and this is what that means: nothing above is granted on Free. If you want to use something you made there, move to a paid plan and generate it again.
• You may not resell the Output as synthetic speech in its own right.Selling or publishing it as a voice dataset, a voice model, a stock library of generated takes, or as training input for somebody else’s speech model is outside this licence. Selling the ad, the audiobook or the podcast the audio is part of is exactly what it is for.
• The licence is conditional on your holding the underlying voice rights described above. Nothing we grant can give you rights in a person’s voice that they did not give you first.

Paid plans are prepaid. Each charge is settled from your account credit first, if you have any, and only the remainder goes to the card on file. Card details go straight from your browser to Authorize.Net and are tokenised there; we never receive or store a card number, and the only card facts kept in our database are the brand, the last four digits and the expiry month and year, so the console can tell you which card is on file.

Account credit

• Account credit is a balance in US dollars held on the workspace. It arrives from a promotional code — entered when you sign up, or entered later by the workspace owner under Account credit in billing settings — or because we put it there. It is not a character allowance and it is not a discount on a future price.
• It is applied automatically to the next charge — a renewal or a prorated upgrade — and only the difference is taken from your card. If the balance covers a charge in full, no card is needed for that charge. Whatever is left rolls forward.
• Credit does not expire. A promotional code has an end date; the credit it granted does not. Every grant and every spend is listed with the balance after it, under Account credit in billing settings.
• It has no cash value and is not refundable. We do not convert it to money, and it does not transfer between workspaces. We may remove unspent credit that was granted in error, or granted by a promotional code that was used in a way it was not offered for — credit you have already spent is not clawed back. Withdrawing a promotional code takes back only what that code granted and you have not spent; credit on the same workspace from any other source is not touched.
• Where this deployment is not configured to take payments at all, there is nothing for credit to offset. The balance is still yours and still shown; it simply cannot be spent until payments are enabled, and the billing page says so rather than implying otherwise.

Periods and renewal

• A billing period is 30 days, not a calendar month. It starts when you subscribe and renews automatically on the same 30-day cycle until cancelled.
• Prices are in US dollars and are charged in US dollars. If your card is issued in another currency, your bank sets the exchange rate and may add its own fee; we neither see nor control either.
• Prices are exclusive of tax. The amount charged is the plan price as listed, with no tax added — the service does not calculate one. Where sales, use or a similar tax is due on your purchase and we are not collecting it, accounting for it is yours to do. If we become required to collect tax on your subscription we will add it to the charge and tell you before the first charge that includes it.

Changing plan mid-period

• Upgrading charges the difference between the two plan prices, scaled to the part of the period still ahead. The new allowance applies immediately.
• Downgrading is not refunded. You keep the plan you paid for — including its allowance — until the end of the current period, and the lower price takes effect at the next renewal. Nothing is credited for the unused part of the period, because the full allowance was available to you throughout it.

Cancelling

You can cancel at any time from the billing settings. Cancelling stops the next renewal; it does not end the current period. You keep the paid plan and its allowance until the period ends, and the workspace then moves to the Free plan. There is no cancellation fee.

Failed payments

If a charge is declined we retry it up to three more times, one, three and five days later. Your service keeps running throughout — a declined card is a payment problem, not a trust problem. If every attempt fails, the workspace drops to the Free plan and its allowance drops with it. Paid content already generated is not deleted.

Refunds

The service issues no automatic refunds. There is no self-serve refund path and no proration on the way down. If you were billed in error, or something went wrong that we should stand behind, email support@workflowcorp.com and a person will decide; we can reverse a charge at the gateway when that is the right answer.

Where a charge was settled partly from account credit, only the part your card actually paid can be reversed at the gateway — the rest is returned to your credit balance by hand, by the same person. Your payment history shows both figures on every charge for exactly this reason, so neither of us has to work out the split from a total.

There is no cooling-off period, because US law does not give you one here and we are not inventing one. There is no general right to withdraw from a digital service in the United States once it has been supplied, and we do not offer a trial-period money-back promise on top of that. What you have instead is the cancellation above: you can stop the next renewal at any moment, from the console, without asking us, and you keep what you paid for until the period ends. If you are a consumer somewhere whose law does give you a withdrawal or refund right that cannot be contracted away, this paragraph does not take it away — tell us and we will honour it.

We aim to keep Voice Buddy available but do not promise that it will be uninterrupted or error-free. No plan carries an uptime commitment, a service credit or a support response time. There is no service level agreement, on any plan, at any price on the pricing page, and no plan carries a support response time. If you need contractual availability or a response time, that is a separate written agreement — email support@workflowcorp.com — and until one is signed, nothing here or on the pricing page creates one.

• “Priority in the render queue” is one specific, bounded thing, and it is worth reading before you buy it for something it is not. Where a plan lists it, work this workspace puts on the background queue is ordered as though it had been waiting up to 8 minutes longer than it actually has. That is the entire mechanism: a head start, not a separate lane and not reserved capacity. Two consequences we would rather state than have you infer. First, it orders only the work that goes through that queue — batch and long-form projects, mixes, dubs, multivoice documents and voice previews — while a render started in the studio and a call to the text-to-speech API run inside their own request and never queue at all, so on those a paid plan is not faster. Second, the bound is deliberate and cuts both ways: no plan, at any price, overtakes work that was queued more than 8 minutes before it, which is what stops a free workspace waiting behind an endless paid one. Nothing about it changes how fast a render itself takes, and it is not a support commitment.
• A slow render is not an outage. Most requests start rendering straight away. When demand scales the speech engine onto an additional instance, that request waits through a model load of roughly 30 to 90 seconds before it produces anything — which is the engine starting, not a fault. The client deliberately waits up to 15 minutes before giving up.
• Some surfaces depend on a third-party provider being configured — sound-effect generation, dubbing translation, the script writer, the assistant, voice design, live conversation with a voice agent, generated music beds, the generated stills and backgrounds in the video maker, the traffic and weather feeds and the Freesound library among them. Where one is not configured, the control that needs it is disabled with the reason on it, or the job stops and reports the reason; nothing is quietly rerouted elsewhere. We do not promise any particular third-party provider will remain available.
• A voice agent, and anything you connect a telephone number to, depends on two outside parties rather than one, and neither of them is us. A live conversation is held by an outside conversation provider — our own speech engine has no way to hold one — and the audio travels between whoever is speaking and that provider directly rather than through this service. A telephone number stays with your own carrier: you keep the account, you keep paying them, and what you give us is permission to answer the number by pointing its call webhook at us. So an agent stops working if either that provider or your carrier does, and neither outage is one we can fix or are promising against. It also means the minutes you spend with your carrier are billed by your carrier and are nothing to do with the allowance described above, which is what the conversation provider charges us.
• Outbound email is the exception, and we would rather say so than have you discover it. The service sends exactly two kinds of message — workspace invitations and password resets — and with no mail provider configured both are written to our server log instead of being sent. Nothing is disabled, and the person who triggered the message is told it succeeded, so an invitee simply never receives it. If somebody you invited never heard from us, that is the first thing to suspect. Nothing about billing is emailed at all: a failed payment appears as a banner on the billing page for members who can manage billing, alongside the payment history and the date the workspace would drop to Free. That page is where to look, not your inbox.
• We may add, change or remove features. If a change materially reduces what a paid plan does, we will tell you at least 30 days before it takes effect, by the same route as a material change to these terms.

We may suspend or terminate an account or a workspace if you breach these terms or the Acceptable Use Policy, if we are required to by law, or if your use threatens the security or stability of the service. For serious abuse — cloning a voice you have no right to, fraud, impersonation, sexual content involving a real person’s voice without consent — we will act without notice.

You can stop using Voice Buddy at any time. An owner can delete a workspace from its settings, which permanently removes its voices, generation history, projects, API keys and webhooks and queues the deletion of its stored audio. You can delete your own account from Settings — with one restriction: if you are the last owner of a workspace that still has other members, hand ownership over first, so that nobody is left holding a workspace they cannot bill or manage.

What survives a deletion, and for how long, is set out in the Privacy Policy. Read it before you rely on deletion — it is more specific than this paragraph, and some things genuinely do outlive the account.

Export anything you want to keep before you delete a workspace. We are not obliged to retain a copy for you.

The service is provided “as is”. To the fullest extent permitted by law we disclaim all implied warranties, including merchantability, fitness for a particular purpose and non-infringement.

Synthetic speech is a statistical process. It mispronounces names, misreads numbers and dates, and sometimes produces a take that is simply wrong. Transcription and translation are approximate. Listen to anything before you publish it, and do not use output unchecked where accuracy matters — medical, legal, financial, safety or emergency contexts included.

Nothing in this section limits liability that cannot be limited by law, including for death or personal injury caused by negligence, or for fraud.

Subject to the paragraph above, our total liability to you arising out of or relating to these terms — in contract, in tort, or on any other basis, taken together across every claim — is capped at the greater of the fees you paid us for Voice Buddy in the twelve months before the claim arose, or USD 100. We are not liable for indirect or consequential loss, for loss of profit, revenue, business, goodwill or anticipated savings, or for loss of or corruption of data, however caused.

That cap is small because the price is small: a workspace on the Free plan has paid nothing, and the paid plans run from a few dollars a month upward. If what you are making carries risk out of proportion to what you pay us — a national campaign, a broadcast obligation, a contract with penalties in it — the protection that fits is your own insurance and your own review of the audio before it airs, not a number in this document. We would rather tell you that than write a cap you would only discover was low when you needed it.

This cap limits what we owe you. It does not limit what you owe us under the indemnity below, and it does not apply to fees you have agreed to pay for the service.

You will defend and indemnify us against claims, damages and reasonable costs arising from your Input, your Output, or your breach of these terms — most obviously a claim by a person whose voice was cloned without their consent, or by a rights holder in audio you uploaded.

How that works in practice:

• We tell you promptly. We will notify you in writing once we are aware of a claim covered by this section, without undue delay. Late notice does not release you, except to the extent the delay actually prejudiced the defence.
• You run the defence. You take sole control of it, with counsel of your choosing. We may participate with our own counsel at our own cost, and you will keep us reasonably informed.
• No settlement that binds us without our written consent. You may settle freely where the settlement is money you pay, releases us fully, and admits no fault on our part. Anything else — an admission, an injunction, an ongoing obligation on us — needs our agreement, which we will not unreasonably withhold.
• We co-operate; you pay for it. We will give you the records, the logs and the account history the defence reasonably needs, and you reimburse our out-of-pocket costs of doing so.

This indemnity is not subject to the liability cap above. That is deliberate and it is the honest shape of the risk: the claim we are most likely to face is one brought by a person whose voice was cloned without their permission, we have no way to check that permission before the render, and the customer who uploaded the clip is the only party who ever knew the answer.

We may update these terms. For a material change — at least 30 days’ notice before it takes effect, given two ways: by publishing the new version on this page with its new effective date at the top, and by emailing workspace owners. The date on this page is the one that governs. Email is the courtesy and the page is the record — mail can bounce, be filtered, or (as the availability section admits) not be sent at all, and none of that should decide when a contract changed.

A change that is not material — a clarification, a corrected cross-reference, a new phone number — takes effect when it is published. Continuing to use Voice Buddy after the effective date means you accept the new version. If you do not accept it, cancel before that date: cancellation stops the next renewal and the period you already paid for runs out as normal.

These terms are governed by the law of the State of Iowa, USA, without regard to its conflict-of-laws rules, and the state and federal courts located in Scott County, Iowa have exclusive jurisdiction over any dispute arising out of or relating to them. Each of us consents to personal jurisdiction there. Iowa is simply where Workflow Corporation is — Davenport is in Scott County — rather than a forum chosen for its case law.

There is no mandatory arbitration clause and no class-action waiver in this agreement. We are not asking you to give up a court or a jury, and we would rather say that plainly than bury the opposite in a paragraph nobody finishes.

Before either of us files anything, tell the other side what is wrong. Email support@workflowcorp.com with what happened and what you want; we will respond, and most of what reaches a filing could have been settled by somebody reading it first. If it is not resolved within 30 days of that email, either of us is free to go to court.

If you are a consumer resident somewhere whose law gives you the right to bring proceedings in your local courts, or applies consumer protections that cannot be contracted away, this section does not take that away from you.

• These terms, plus the Acceptable Use Policy and the Voice Cloning and Biometric Data Notice, are the whole agreement between us about Voice Buddy.
• If a provision is unenforceable, the rest continues to apply.
• Not enforcing a term once does not waive it.
• You may not assign this agreement without our consent. We may assign it to a successor of the business.

Questions about these terms, and formal notices under them, go to support@workflowcorp.com — mark a legal notice for the attention of the legal team in the subject line, which is what routes it. There is one address on purpose: a legal@ or a notices@ that nobody watches is worse than a shared inbox that is read. Post reaches us at Workflow Corporation, 210 Emerson Pl, Suite 300, Davenport, IA 52801, USA, and the phone is +1 (563) 275-6409.

Support is available at any hour. The office is open Monday to Friday, 8:00 AM to 6:00 PM Central, and closed at weekends, so anything needing a decision rather than an answer — a notice, a refund, an appeal — is dealt with on a business day.