Acceptable Use Policy

Audexum turns text into speech, speech into text, dubs video, builds audiobooks and lets you clone a voice and generate images and video. The rules below come down to these:

• Clone your own voice, or a voice you have written permission to use. Nobody else's.
• Only upload text, audio, video and books you have the rights to.
• Do not use synthetic speech to deceive people about who is speaking.
• Audio made with V2 Multilingual must be labelled as AI-generated wherever you publish it (section 13A).
• Image and video tools: adults only, no sexual content, no real person's face or voice without their consent, and never remove the AI marks (section 13B).
• Nothing illegal, and nothing aimed at children.

If we act against your account we will tell you why, your unused credits stay yours unless the breach was serious, and you can appeal. The numbered sections below are the binding part.

This Acceptable Use Policy (the "Policy") is part of your agreement with Audexum and binds everyone who uses the Service — the website, the Studio, the Android and Windows apps, the API and the ElevenLabs-compatible API, whether on a free or a paid plan. It applies to what you submit (your Input), what the Service produces (the Output), and what you do with the Output afterwards.

You are responsible for everything done through your account and your API keys. If you let other people use the Service through your account, product or website, you are responsible for making sure they follow this Policy too.

Some of the AI models behind the Service carry their own licence restrictions. Where a model licence is stricter than this Policy, the stricter terms apply to Output from that model.

The Service does offer custom voice cloning on paid plans. To enroll a voice you must submit a recording of the speaker reading aloud a consent sentence we generate. We check the recording with our speech-to-text engine, keep the resulting speaker embedding and a record of the consent, and keep an encrypted copy of the recording for up to 12 months, as described in the Privacy Policy.

You may only enroll:

• your own voice; or
• the voice of a person who has given you documented, specific, informed permission to clone their voice for the uses you intend, and who is the person actually reading the consent sentence; or
• a voice you have acquired the rights to by written agreement — including the rights of the person who recorded it, not only the rights of a studio or a label.

You must not:

• enroll the voice of a politician, official, celebrity, journalist or any other public figure without their permission — being publicly known is not permission;
• enroll a voice from a podcast, film, broadcast, video, phone call, voice message, dataset or any other recording you did not make with the speaker's agreement for this purpose;
• enroll the voice of a deceased person without the documented agreement of whoever holds those rights under the applicable law;
• enroll the voice of a child;
• have someone else read the consent sentence in place of the person whose voice is being cloned, or splice a consent recording onto a clip of a different speaker;
• use a cloned voice to make a real person appear to say something they did not say;
• share, sell or transfer a cloned voice to anyone who does not hold the same permission you do.

The consent check is a barrier, not a guarantee. It confirms that someone read the sentence; it cannot confirm that the person had the right to. Passing the check is not permission and is not a defence. The obligation to hold real permission is yours, and you warrant to us that you hold it for every voice you enroll. We keep the consent record so that a complaint about a cloned voice can be investigated.

If the permission behind a voice is withdrawn or turns out never to have existed, stop using that voice and delete it. Tell us at /report if the Output has already been published.

You must not use the Service to:

• generate audio that purports to be, or could reasonably be mistaken for, a real person speaking, without that person's permission — with a cloned voice, a preset voice or any other means;
• put words in a real person's mouth, whether or not the voice sounds like theirs;
• impersonate Audexum, our staff, a company, a public body, a bank, or any other organisation;
• present Output as a human recording where a listener would care about the difference — for example customer-service calls, news, medical, legal or financial content, or content aimed at children;
• remove, alter, obscure or defeat the watermark or any provenance marker we embed in the Output;
• fail to disclose that audio is AI-generated where the law requires you to, including Article 50 of the EU AI Act.

Disclosure to your own listeners is your responsibility and cannot be delegated to us. What we do and do not mark is described at /ai-disclosure.

You must not use the Service to:

• generate audio of a candidate, party figure, election official or head of state that could mislead a listener as to authenticity, at any time and whether or not an election is running;
• produce content designed to suppress turnout, misstate how, when or where to vote, or fabricate a concession, endorsement or withdrawal;
• impersonate an election authority, court, regulator, police force, tax authority, or any other public institution;
• run coordinated inauthentic campaigns — many voices or personas presented as independent people when they are not.

Dubbing, audiobook generation, dictation and batch synthesis all take content you supply. You must hold the rights to everything you upload — the text, the book, the script, the audio track, the video and any voices recorded in it.

In particular, you must not upload or publish:

• books, articles, scripts, lyrics or subtitles you do not own or have a licence to use for this purpose — a purchased copy is not a licence to publish a synthetic reading of it;
• video or audio you did not create and are not licensed to dub, redistribute or make derivative works from;
• recordings of people who did not agree to be recorded, dubbed or synthesised;
• content that infringes a trade mark, a database right, a personality or publicity right, or a right of privacy;
• material obtained by circumventing a paywall, a technical protection measure, or another platform's terms of service;
• transcripts from the free YouTube transcript tool in a way the video's owner has not allowed, such as republishing a whole transcript as your own work.

The same applies to anything you make publicly available through the Service, such as share clips and podcast feeds. Publishing through us makes it visible to people who are not our users, so treat it as publication anywhere else: it must be lawful in every country where it can be heard, and it must not require a licence you do not hold. Rights-holders can ask us to take content down — see section 15.

You must not use the Service to generate, process, host or distribute material that sexualises any person who is, or appears to be, a child. This covers child sexual abuse material in audio form, sexual roleplay involving minors, grooming scripts, and uploads of such material for dubbing or transcription.

You must not use the Service to deceive, manipulate, sexually solicit or exploit children, and you must not clone a child's voice.

Accounts used for this are terminated permanently. We preserve the associated records for as long as the law allows and report to the competent authorities in Bulgaria. This is the one category where an appeal will not restore the account.

You must not use the Service for, or to help anyone else carry out:

• phishing, vishing or any other social-engineering scheme;
• voice-based scams targeting individuals — fake-relative and fake-kidnapping calls, fake bank or courier calls, fake authority calls;
• business email or voice compromise, including cloned executives instructing a payment;
• fake reviews, fake testimonials, fake interviews or fabricated news audio;
• market manipulation, pump-and-dump schemes or insider dealing;
• unsolicited automated calls, robocalls or IVR campaigns that break the law where the recipient is — including the EU ePrivacy rules, the UK PECR and the US TCPA;
• collecting personal data or payment details under a false pretence.

Voice fraud is the highest-risk use of this Service. We report credible instances to law enforcement.

You must not generate, upload or distribute content that:

• harasses, threatens, stalks or intimidates a person, or is used in a campaign against them;
• publishes someone's private information, address or contact details without their agreement (doxxing);
• makes false factual statements that damage a person's or an organisation's reputation;
• is non-consensual sexual content, or sexual content depicting a real identifiable person without their agreement;
• incites violence, terrorism or genocide, or supports a proscribed organisation;
• encourages suicide, self-harm or eating disorders;
• attacks people on the basis of race, ethnicity, national origin, religion, disability, age, sex, gender identity or sexual orientation;
• contains malware, malicious code or an executable payload, in the Input or attached to it;
• is otherwise unlawful where you are, where the subject is, or where the Output will be heard.

Where the Service can capture microphone or system audio, it records everyone in the conversation, not only you. Those people are not our users, cannot see your screen, and any prompt or confirmation the app shows is shown to you alone — it is not their consent and we cannot obtain it for you. You must not:

• record a conversation without the consent of every participant, where the law applicable to any of them requires it;
• record anyone covertly, or where they reasonably expect not to be recorded;
• keep recording after someone objects, or keep a recording or transcript they have asked you to delete;
• record someone in order to monitor them, profile them, or build a voice sample of them.

You must not submit Input that:

• contains other people's personal data without a lawful basis under the GDPR;
• contains sensitive data — health, biometric, sexual, religious, political, trade-union or precise location data — or financial-account, payment-card or government-identifier data, without a proper basis and appropriate safeguards;
• you obtained unlawfully, including from a data breach or from stolen credentials.

If you are processing personal data through the Service on behalf of your own customers, see /dpa. How we handle data is set out at /privacy.

You must not:

• create multiple accounts to get around free-tier limits, a failed payment or a suspension;
• share, sell or publish your API key, or embed it where end users can extract it;
• scrape, crawl, mirror or bulk-extract data, voice embeddings or model weights from the Service;
• reverse-engineer or decompile the Service or the models behind it, except as far as the law expressly allows and does not let us restrict;
• use the Service, the Input or the Output to build or train a competing speech synthesis, voice cloning or voice conversion system;
• bypass quotas, rate limits, authentication, payment or content-filtering controls;
• disrupt the Service, including by sending traffic far above what your plan allows;
• republish or resell the Service as if it were your own product;
• abuse the referral bonus — for example by referring yourself or creating accounts to collect it; see Terms, section 12.

Good-faith security research is welcome and is not a breach of this Policy, provided you stay inside the rules we publish at /.well-known/security.txt:

• report what you find to [email protected] with steps to reproduce, and give us reasonable time to fix it before publishing;
• do not test against other users' data or accounts — use your own;
• do not run denial-of-service or load tests without our prior written approval;
• do not access, modify, delete or retain data that is not yours, and do not use a finding to gain anything beyond proving it;
• do not use social engineering, physical intrusion, or attacks against our staff or our suppliers.

We aim to acknowledge security reports within 48 hours. We will not pursue you for research that stays within these rules. We do not currently run a paid bug-bounty programme, so please do not expect a reward. Testing that goes outside these rules is covered by section 11 instead.

The Service is not designed or tested for emergency communications, life-support or medical decision systems, vehicle, aviation or maritime control, nuclear facilities, or any other setting where an error in the audio could cause death, injury or serious damage. Do not deploy it there.

Do not use Output to deliver automated decisions with legal or similarly significant effects on people — employment, credit, housing, insurance, benefits — without human review and a way for the person to contest it. Do not hold Output out as licensed medical, legal or financial advice. Do not target vulnerable people in ways that exploit their situation.

The V2 Multilingual voice tier runs on a third-party model (Supertonic-3, under the OpenRAIL-M licence) whose licence requires us to pass its use restrictions on to you. They apply to any Output generated on V2 Multilingual, wherever you generate it: the Studio, the API, batch, audiobooks or any integration. They are in addition to everything else in this Policy. They apply to accounts opened on or after 5 October 2026. You must not use V2 Multilingual, or its Output:

• in any way that breaks any applicable law or regulation;
• to exploit or harm minors, or try to, in any way;
• to create or spread verifiably false information meant to harm others;
• to create or spread personal information that can be used to harm someone;
• to publish or place Output anywhere without clearly stating that it is AI-generated (machine-generated) speech;
• to defame, disparage or harass anyone;
• to impersonate anyone, or try to, without their consent, including deepfakes;
• for fully automated decisions that harm a person's legal rights or create or change a binding obligation;
• to discriminate against or harm people or groups based on their online or offline behaviour or their known or predicted personal or personality traits;
• to exploit the weaknesses of a group of people because of their age or social, physical or mental characteristics, in a way that changes how they behave and is likely to cause them or someone else physical or psychological harm;
• to discriminate against people or groups based on legally protected characteristics;
• to give medical advice or interpret medical results;
• for the administration of justice, law enforcement, immigration or asylum, such as predicting whether someone will commit fraud or a crime.

The AI-generated label rule applies to V2 Multilingual only. On the other tiers you must still disclose AI-generated audio where section 3 or the law requires it.

Voices served by hosted providers. Some voice tiers, and some voices while our own GPUs are busy, run on models hosted by other providers through OpenRouter, among them Google (Gemini), xAI, Fish Audio and hexgrad (Kokoro). Their usage policies apply to that Output as well, including Google's Generative AI Prohibited Use Policy and xAI's Acceptable Use Policy. Nothing they prohibit is allowed here either.

The image and video tools (Terms, section 11A) run on third-party models whose providers require us to pass their rules on to you. These rules apply to every prompt, uploaded image and result on those tools, in addition to everything else in this Policy. They apply to you from the moment you turn the tools on.

Who may use them. You must not use the image and video tools if:

• you are under 18;
• you are in, or acting for, a country, person or entity under the sanctions in section 14;
• we have turned the tools off for your account, or terminated an earlier account of yours.

Content you must not create, upload or share. You must not use the tools, or their results, to create or spread:

• any sexual content involving a person who is, or appears to be, under 18, or any content that sexualises, grooms, exploits or harms children, including sextortion material. Zero tolerance: section 6 applies and we report it to the authorities;
• obscene or harmful material aimed at or sent to minors;
• sexually explicit or pornographic content of anyone, or content made for sexual gratification. The tools are for non-sexual content only;
• intimate, sexual, nude or undressed imagery of a real person without their consent (non-consensual intimate imagery), whether or not it looks real;
• a realistic image or video of a real, identifiable person, living or dead, without their consent, or one that shows them in a degrading, defamatory, fraudulent, misleading or abusive way;
• a result that sounds like a real, identifiable person without their consent. Every audio clip you upload as a reference must be your own voice, or a voice whose speaker has agreed to this use;
• depictions or promotion of human trafficking, sexual violence or exploitation;
• gratuitous violence, extreme gore, torture, mutilation or animal abuse;
• content that promotes or facilitates self-harm, suicide or eating disorders;
• terrorist or violent extremist content, or content that recruits for, incites or supports such groups;
• threats, harassment, bullying, intimidation or incitement to violence against anyone;
• hate speech, slurs, or content that dehumanises people or argues for discrimination against them because of a protected characteristic;
• profanity or imagery meant to shock, disturb or harass;
• disinformation, including fake images or video of real events, people or public figures meant to mislead, and content likely to harm public health, public safety, elections or civic participation;
• content for political campaigning, electoral advertising or voter suppression that uses a realistic depiction of a real person;
• content that falsely presents events, things or people in a way likely to deceive, without the disclosure the law or this Policy requires;
• content that infringes copyright, trade marks, privacy, publicity, personality or image rights, or that copies or closely imitates a specific protected work, character or brand you have no rights to;
• content for illegal or regulated goods or activities, or any other crime.

Deception and fraud. You must not use the tools or their results for:

• scams, fraud, phishing, fake identity documents or any scheme to deceive people or obtain an unfair advantage;
• impersonating a real person, a company or a public body, including deepfakes that use someone's name, face, body, voice or likeness without their consent;
• manipulative content designed to distort people's decisions in a way likely to harm them;
• misleading claims of expertise in health, law, finance or government services;
• spam or unsolicited bulk content.

Provenance and labelling. You must:

• not remove, alter, hide, suppress or defeat the provenance marks, Content Credentials (C2PA), watermarks or AI-generated metadata in a result, or make or share a tool for doing so;
• not claim or imply that a result was made by a human without AI, or that it is a real photograph or recording;
• clearly disclose that a result is AI-generated or manipulated when you publish one that shows real people, places or events in a realistic way, as Article 50(4) of the EU AI Act requires of you as the deployer;
• keep that disclosure attached when you re-post, crop, edit or syndicate the result.

Personal data and surveillance. You must not use the tools to:

• upload an image of a recognisable person without their consent, or of a child unless you are their parent or guardian and the use is lawful and non-sexual;
• collect, process or disclose other people's personal data, health data, financial or payment data, or sensitive data such as racial or ethnic origin, political opinions, religious beliefs, sexual orientation or trade-union membership;
• process biometric data, recognise or identify people from their face or body, build a face database, or infer sensitive traits about a person;
• track, monitor or profile people without their consent;
• analyse metadata or model behaviour to work out someone's identity, personal information or the origin of content;
• support military, surveillance or law-enforcement purposes, any system classed as high-risk under the EU AI Act, or any practice Article 5 of the AI Act prohibits;
• make fully automated decisions that affect a person's legal rights or create binding obligations, or discriminate against people in employment, credit, housing, education or other essential services.

The engines themselves. You must not:

• try to reach the engines other than through the web application, automate or script the tools, or let anyone else use them through your account;
• resell, sublicense or offer access to the tools, or use them to build a similar or competing service;
• use the results, or the tools, to train, fine-tune or improve any AI model that competes with the engines or with us;
• try to extract model weights, source code, training data or the system behind an engine, or reproduce its training data;
• circumvent or test the limits of safety checks, content filters or guardrails, for example by rewording a refused prompt to get the same result;
• scrape or bulk-download from the tools, overload them, introduce malware, or attempt unauthorised access.

Model providers' policies. You also agree to follow the usage policies that the providers and developers of the models behind the tools publish, as they apply to the engine you use. They are listed, with a link to each one the provider publishes, at /ai-disclosure. Where one of them is stricter than this Policy, the stricter rule applies to results from that engine.

Enforcement. Breaking these rules is a breach of this Policy under section 16. In addition, repeated blocked requests may pause the tools on your account automatically, for example for 24 hours after three blocks in a day, and any attempt to create sexual content involving minors turns them off permanently and is handled under section 6.

You must not use the Service, or the Output, in or for the benefit of any country, person or entity subject to comprehensive EU, UK, US or UN sanctions, or in any way that breaches EU dual-use export controls under Regulation (EU) 2021/821.

Use /report to tell us about anything on or from the Service that breaks this Policy or the law — a cloned voice that is yours and that you did not agree to, an impersonation, a scam, harassment, illegal content, or a share clip that infringes your rights. You do not need an account to report. You can also write to [email protected], or to [email protected] for copyright and other rights claims.

Tell us where the content is (a URL if you have one), what is wrong with it, why you believe you are right, and how to reach you. Rights claims should identify the work and your relationship to it. Reports made in bad faith, or in bulk without checking, may cost you access to the reporting route.

We assess reports and act where we agree. If we remove or restrict content, we tell the person who posted it what we did and why, and both of you can use the appeal route in section 18. Security vulnerabilities go to [email protected] instead — see section 12.

What we do depends on how serious the breach is, whether it looks deliberate, and whether it has happened before. In order of severity:

We normally warn first. We may skip straight to suspension or termination where the breach is serious, where waiting would cause harm, or where the law requires us to act. Whenever we suspend, restrict or terminate, we tell you what we did, on what ground, and how to appeal — unless a law or a live investigation stops us from saying.

Your credits. We do not forfeit your prepaid credits, and there is no clause anywhere in our terms that lets us. Suspension does not take them away, and neither does termination — whatever the ground, including a material breach of this Policy. Your unused balance is handled under /refunds in every case. If your breach causes us actual loss we may claim compensation for it under the law, but that is a claim we would have to make and prove, not a balance we help ourselves to.

Nothing here limits your statutory rights as a consumer, or our right to claim compensation the law allows for loss you actually cause us.

We keep a record of enforcement actions against an account. If, after being told at least twice that specific content or conduct infringes someone's rights or breaks this Policy, you do it again, we treat you as a repeat infringer and terminate the account. A single serious infringement can be enough on its own.

Before we terminate on this ground we give you notice and a chance to answer, unless the content is child sexual abuse material or presents an immediate risk to someone. We disregard notices we assessed and rejected, and notices withdrawn by the person who sent them.

Opening a new account to get around a termination is itself a material breach.

If we warn you, remove your content, suspend a feature, suspend your account or terminate it, you can appeal. Email [email protected] within six months, quoting the account email and what we did. People who reported content can appeal a decision not to act, on the same route.

A person reviews the appeal — not an automated system. We aim to answer within a reasonable time and will tell you the outcome and the reason for it. If we got it wrong we restore the account, the content and any credits affected. Appeals about child sexual abuse material are reviewed but will not restore the account.

Appealing to us does not remove your right to go to a court or to a consumer authority — for Bulgaria, the Commission for Consumer Protection (Комисия за защита на потребителите). Where you live in the EU, you can always bring proceedings in the courts of your own country.

We may update this Policy as the product, the threats or the law change. The current version is always the one at /aup, with its version number and date at the top. We announce material changes on the site before they take effect and, where we have your email and the change matters to you, by email. If you do not accept a change, stop using the Service and close your account; /refunds covers what happens to an unused balance.

• Report abuse or illegal content: /report or [email protected]
• Rights claims, appeals, legal notices: [email protected]
• Security reports: [email protected]
• Privacy and data-protection requests: [email protected]

The Service is operated by Audexum, established in Bulgaria. Full operator details, including the status of the company registration, are at /imprint. This Policy sits alongside /terms, which governs the rest of the relationship.