Custom Voice Feature Terms of Service

These Custom Voice Feature Terms of Service (these “Terms”) are specific service terms under the GPTBots.ai User Service Agreement (the “Main Agreement”) and govern your use of the custom voice feature service on the GPTBots Platform (the “Feature” or the “Service”). These Terms constitute an agreement between you and METAVERSE CLOUD PTE. LTD., GPTBots Hong Kong Limited, METAVERSE CLOUD SDN. BHD., and their other affiliates (“we,” “us,” or “our”). Matters not addressed in these Terms are governed by the Main Agreement. In the event of any inconsistency between these Terms and the Main Agreement, these Terms prevail with respect to the Feature. Provisions that limit or exclude liability or otherwise materially affect your rights and interests are highlighted in bold for your particular attention.

You may accept these Terms only if you are authorized to do so on behalf of your organization. By checking the acceptance box, you confirm that you have read, understood, and agreed to all of these Terms. If you do not agree, do not use the Feature. Acceptance of these Terms does not replace any separate authorization or consent required from a voice owner for the owner’s voice, recordings, performances, or processing of personal data. We will retain records of the version of these Terms, the time of your acceptance by checkbox, the organization, and the person taking the action.

Definitions

Term	Meaning
Feature	The custom voice feature service provided on the GPTBots Platform, available in two modes: GPTBots Credential Mode, which allows you to train a voice through the GPTBots Platform by uploading voice materials to the Platform and submitting speech-synthesis requests through the Platform to have a third-party voice model service train on those materials and generate and return a successfully trained voice (a “Platform-Trained Voice”); and Own Credential Mode, which allows you to bind to the GPTBots Platform a voice that you have already successfully trained with a third-party voice model service provider (an “Own Voice”), and to use the bound Own Voice to submit speech-synthesis requests through the Platform to the third-party voice model service and receive synthesized speech.
GPTBots Credential Mode	You do not need to register an account with a third-party voice model service provider. You may upload voice materials directly to the GPTBots Platform and submit speech-synthesis requests through the Platform to have a third-party voice model service train on those materials and generate and return a successfully trained voice. Each such trained voice consumes a voice quota included in your paid subscription plan.
Own Credential Mode	You have your own account and credentials (KEY) with a third-party voice model service provider. You must bind your own KEY and Own Voice on the GPTBots Platform. When you use the Own Voice through the Platform to submit speech-synthesis requests to the third-party voice model service and receive synthesized speech, the corresponding amount will be deducted from your balance with that provider, and we will charge the applicable GPTBots credits as a service fee.
Voice Materials	Audio files recorded or uploaded by you to the GPTBots Platform for voice training.
Own Voice	A voice that you have successfully trained with a third-party voice model service provider.
Custom Voice	A speech-model resource generated by training on Voice Materials (including its identifier, SpeakerID), or an Own Voice bound by you.
Voice Owner	The natural person whose voice is recorded in the Voice Materials.
Synthesized Content	Speech audio files and other output generated and returned through the Feature when you use a Custom Voice to call a third-party voice model service.
Third-Party Voice Model Service Provider	A third-party provider that supplies voice-training and speech-synthesis capabilities for the Feature. The names, terms of service, and usage policies of the specific third-party voice model service providers integrated with the GPTBots Platform are available at https://www.gptbots.ai/zh_CN/docs/getting-started/model-list.

1. Authorization and Warranties

1.1 You represent and warrant that, with respect to the Voice Materials you upload to the GPTBots Platform and any Own Voice you bind on the GPTBots Platform, one of the following applies:

(1) You are the Voice Owner; or

(2) You have obtained written or other demonstrable, specific, and valid authorization or consent from the Voice Owner, and have the right, within the scope of the rights actually held by the relevant rights holders and the authorization granted, to permit you, us and our affiliates, and the third-party voice model service providers necessary to provide the Feature, within the agreed scope, to collect, transmit, store, process, and reproduce the Voice Materials and the performances, recordings, and other protected content contained therein; create acoustic representations/Custom Voices; generate Synthesized Content; and provide the Feature to you. The Voice Owner’s authorization must clearly state the commercial purposes, service providers, purposes, retention/deletion restrictions, relevant processing locations or cross-border processing, withdrawal channels, and consequences of withdrawal. You may not infer that the Voice Owner has authorized such use solely from general terms in an employment or service contract. You must ensure that you have obtained or otherwise hold the copyright and performers’ rights required for the processing described in this clause and that you comply with applicable law. You may not upload a minor’s voice unless you have obtained authorization from the minor’s guardian that has been separately verified.

Your use of the audio files constituting Voice Materials that you input, upload, or provide to the GPTBots Platform, and of any Own Voice you have bound on the GPTBots Platform—including, without limitation, collection, analysis, generation, and synthesis of the voice—will not infringe any third party’s intellectual property rights, including copyright, portrait rights, voice rights, privacy rights, reputation rights, rights of publicity, or other lawful rights and interests.

1.2 You must determine whether the Voice Materials you upload to the GPTBots Platform and/or any Own Voice you bind on the Platform constitute personal data or specially protected personal information, taking into account whether the Voice Materials, acoustic characteristics, and associated information can identify an individual, the purposes of processing, and the requirements of applicable law. For processing activities whose purposes and means you actually determine, you must fulfill the obligations imposed by applicable law regarding notice, consent (or another lawful basis), purpose limitation, data minimization, security, retention periods, access/correction, cross-border transfers, and data-breach notifications, and conduct impact assessments where required by law. We will fulfill our separate obligations in accordance with our actual role and applicable law, as described in the GPTBots Platform Privacy Policy. For clarity, the foregoing role determination applies only to processing activities whose purposes and means you actually determine; each party must perform its statutory obligations in accordance with applicable law and its actual role.

1.3 You must retain, for the period during which the Voice Owner’s authorization remains valid and for any additional period necessary and permitted by applicable law, records sufficient to demonstrate the scope of the authorization and how it was obtained. Upon reasonable notice from us, you must provide, within a reasonable time, evidence of the Voice Owner’s authorization directly relevant to the dispute, and may appropriately redact or de-identify unrelated personal information. We may ask you to explain or provide necessary materials to address complaints, audits, or legal compliance, but will limit the scope of collection, protect the supporting materials, and delete them in accordance with applicable retention periods. If you fail to provide valid evidence of the Voice Owner’s authorization, or if there are reasonable grounds to question the evidence you provide, we may take necessary and corresponding measures under Section 7 of these Terms.

1.4 If the Voice Owner withdraws their consent or authorization, you must immediately stop using the relevant Voice Materials or Own Voice for additional training and new speech synthesis, and promptly notify us. In accordance with each party’s actual processing role and applicable law, we will stop the relevant processing, assess the request, and implement deletion or restriction of processing, and will submit corresponding requests to the relevant third-party voice model service provider to the extent permitted by applicable law and technically feasible. If you continue to use the relevant Voice Materials or Own Voice after the Voice Owner has withdrawn their consent or authorization, you will be responsible for any resulting infringement or dispute.

2. Restrictions on Use

2.1 When using the Feature, you and your users (including end users and authorized users) must not use the Feature to:

(1) Infringe the rights of any natural person or entity, including copyright, privacy, voice, publicity, reputation, intellectual property, or other lawful rights and interests;

(2) Facilitate unlawful conduct, for example by providing audio instructions for carrying out illegal activities or manufacturing controlled items;

(3) Abuse or harm minors in any way, including grooming minors or child sexual exploitation;

(4) Process a minor’s voice without valid authorization from the minor’s guardian; engage in other interactions with minors that may result in their exploitation, manipulation, abuse, or harm, or that are prohibited by applicable law; or use the Feature for the purpose of harming minors;

(5) Infringe an individual’s privacy or other rights without the consent or other lawful basis required by law, including:

① collecting, processing, or using personal data or biometric information;

② unlawfully tracking, stalking, monitoring, surveilling, or harassing a natural person;

③ identifying or verifying the identity of a natural person based on personal data;

(6) Engage in fraud, deception, dissemination of false information, or misleading conduct, including passing off synthesized speech as human-created content, generating fake engagement, false advertising, scams, phishing, impersonating another person, deceiving an audience into reasonably believing that the voice is that of a specific real person, or otherwise committing identity theft;

(7) Create, replicate, or use a voice replica identifiable as that of a specific natural person (living or deceased) without appropriate consent, valid authorization, or lawful rights, including unauthorized impersonation, non-consensual intimate voice generation, or deceptive impersonation without disclosure;

(8) Impersonate any natural person (living or deceased) for an unlawful purpose without lawful authorization and clear disclosure;

(9) Delete, remove, obscure, alter, or conceal credentials, watermarks, labels, metadata, other marks, or provenance signals used to identify content generated by a third-party voice model service provider;

(10) Exploit vulnerabilities of an individual or a specific group to cause harm, exploit, or manipulate, including targeting individuals based on age, disability, race, socioeconomic status, or other personal characteristics;

(11) Classify or profile natural persons in a manner that may result in discrimination, unfairness, bias, or other harmful treatment;

(12) Infer sensitive personal attributes, including gender, race, religion, or precise age;

(13) Engage in or assist illegal or high-risk activities, including military operations, weapons development, operation of critical infrastructure, activities involving controlled items, or activities that may result in serious injury or death, particularly scenarios involving autonomous lethal functions without human supervision;

(14) Generate, display, process, or use content prohibited by a third-party voice model service provider’s usage policy, content that may facilitate conduct prohibited by such policy, or content related to matters prohibited by such policy, including, without limitation, violent extremism, terrorism, human trafficking, child sexual abuse/exploitation, violence, self-harm, sexually explicit content, hate, harassment, bullying, intimidation, abuse, insults, or discrimination;

(15) Bypass or circumvent model safety filters, abuse-prevention measures, safety mechanisms, or other security controls;

(16) Guide, instruct, manipulate, incite, or induce a third-party voice model service to engage in conduct prohibited by the applicable third-party voice model service provider’s usage policy;

(17) Sell, rent, or transfer a Custom Voice or its access credentials to any third party outside your organization, or use it to develop or train a voice model that competes with us or a third-party voice model service provider; or

(18) Engage in any other conduct that violates the Main Agreement, these Terms, a third-party voice model service provider’s terms of service or acceptable use policy, or applicable law.

2.2 When providing the Feature or Synthesized Content to your users, you must require your users, in written terms, to comply with the use restrictions in Section 2 of these Terms that apply to their use cases, and implement management measures proportionate to the risks. You are responsible for claims arising within your control as a result of your breach of these Terms or your failure to reasonably manage your users.

3. Use and Labeling of Synthesized Content

3.1 Ownership of Synthesized Content is governed by Section 3.1 of the Main Agreement. If the Main Agreement and these Terms are inconsistent as to the scope of the license to process Voice Materials or the Feature, these Terms prevail. Except for the licenses expressly granted in these Terms, you and the relevant rights holders retain the rights in Voice Materials, Own Voices, and Synthesized Content that each is entitled to under applicable law. These Terms do not mean that you necessarily own all rights in the Voice Materials, the recordings, performances, music, text, or other content contained therein; you grant a license only to the extent that you own the relevant rights or have obtained authorization sufficient to cover the uses described in this section. To provide the Feature to you, you grant us a non-exclusive, worldwide license, valid for the period necessary to provide the Feature, to allow us and our affiliates, third-party voice model service providers, and other service providers engaged by us, to the extent necessary to provide, maintain, protect, troubleshoot, and support the Feature, to receive, host, reproduce, transmit, and technically process the Voice Materials you upload to the GPTBots Platform, any Own Voice bound on the Platform, and related files generated from them; train and generate voices and Synthesized Content; and return Synthesized Content to you. We may grant service providers actually involved in providing the Feature the corresponding necessary rights to use such materials solely for the foregoing purposes. We will not use Voice Materials you upload or Own Voices you bind for our own purposes, independent commercial exploitation, or training our own models, and will be subject to appropriate confidentiality, security, use-restriction, and deletion obligations. You must ensure that use of the Voice Materials you upload to the GPTBots Platform and any Own Voice bound on the Platform for the Feature is supported by the necessary rights, licenses, or other lawful basis you are required to obtain, and that the foregoing license is sufficient for the processing described in these Terms. Your subsequent use of Synthesized Content must also comply with applicable law and the requirements of the relevant third-party voice model service provider and must not infringe third-party rights. This section does not guarantee that Synthesized Content will constitute a copyright-protected work, that you will own its copyright, that it will be exclusive or registrable, or that we guarantee that Synthesized Content will not infringe third-party rights.

3.2 You must clearly and appropriately label Synthesized Content as AI-generated/synthesized in accordance with the laws and regulatory requirements applicable to the place where it is disseminated and the relevant use case. Even if not expressly required by law, you must take reasonable disclosure measures where non-disclosure could materially mislead the audience. You must not delete, alter, or conceal watermarks, source labels, or metadata embedded by us or a third-party voice model service provider as required by law or contract or for security and provenance purposes. We will identify the existence of such labels only where the third-party voice model service provider actually supplies them and it is technically feasible. You must not assume that a third-party watermark will necessarily exist or be guaranteed to be non-removable.

3.3 You must implement reasonable human review and appropriate human oversight based on the purpose, audience, and potential harm of Synthesized Content. If Synthesized Content is used in finance, healthcare, legal matters, emergency notifications, identity verification, or any other context that may affect an individual’s important rights and interests, you must establish stricter review and correction mechanisms. We do not replace your professional judgment and do not guarantee that Synthesized Content is accurate or suitable for a particular purpose.

4. Data Processing

4.1 When you use the Custom Voice Feature, we will collect voice materials recorded by you, voice materials of others that you submit, or Own Voices that you provide, in order to extract voice characteristics and generate a Custom Voice so that you can use that voice to synthesize speech content. Before submitting another person’s Voice Materials or voice, you must ensure that you have given the Voice Owner the processing-purpose disclosures and other notices required by applicable law, and obtained the required consent or established another lawful basis for processing.

4.2 To provide the Feature, we may transfer Voice Materials, the identifier SpeakerID, and necessary metadata to third-party voice model service providers that are necessary to perform the Feature and have been disclosed to you. If the transfer constitutes a transfer of personal data outside Singapore, the organization legally responsible for the overseas transfer must take measures under applicable law to ensure that the overseas recipient provides protection comparable to that under the PDPA, and implement restrictions on purposes, retention, deletion, security, and onward transfers. Each party must fulfill its statutory obligations in accordance with its actual role.

4.3 We will process Voice Materials and Own Voices you provide only to the extent necessary to provide the Feature, maintain and secure it, troubleshoot, investigate abuse, comply with legal obligations, and enforce these Terms. Unless you separately and expressly opt in and the necessary rights/consents have been obtained, we will not use Voice Materials or Own Voices you provide to train our models, nor will we sell them.

4.4 The storage locations, retention periods, security safeguards, and mechanisms for protecting user rights with respect to data generated through your use of the Feature are governed by the GPTBots Platform Privacy Policy and applicable data-processing terms, which you may review before using the Feature. The parties will promptly notify each other of data-security incidents involving personal data related to the Feature and reasonably assist with the investigation.

4.5 If you use a voice you generated to create an Agent on the GPTBots Platform and upload that Agent to a GPTBots Platform workspace so that other members of your organization can use it, before making it available you must clearly inform the Voice Owner that the voice will be made available to other members of your organization through the Agent and describe the scope of their use, and obtain any consent required by applicable law or establish another lawful basis for processing. You must ensure that the authorization covers the relevant use by those third parties.

4.6 For Own Voices, you are responsible for obtaining the right to provide us with the SpeakerID, calling credentials, and related metadata, and for your account, licensing, and fee arrangements with the third-party voice model service provider. We will process the voice information you submit only as necessary to provide integration services. Where we receive or process personal data, access credentials, or information that can identify a voice, we will still take reasonable security measures, restrict access, and delete such information in accordance with the GPTBots Platform Privacy Policy, applicable data-processing terms, and applicable law. You must not submit credentials beyond what is necessary for the integration. We make no commitment regarding the availability of third-party services.

5. Deletion

5.1 You may delete a voice at any time by clicking the delete button on the GPTBots Platform. For technical reasons, a voice that is currently being trained cannot be deleted immediately; deletion will be carried out after training is complete. Until deletion or restriction of processing is complete, you must not continue to call that voice for new speech-synthesis requests.

5.2 Deleting a Platform-Trained Voice has the following consequences:

(1) We will delete the voice record, preview audio, and any remaining original Voice Materials from the GPTBots Platform databases under our control, and stop calling the voice;

(2) Any Agent currently using that voice will stop calling it and automatically switch to the default voice provided by the third-party voice model service provider;

(3) After you click to delete a voice, we will unbind the SpeakerID associated with that voice and stop calling the speech-model resource that was generated and stored by the third-party voice model service provider after the provider trained on your uploaded Voice Materials in response to your speech-synthesis requests. However, we cannot automatically delete that resource through the API. If you wish to exercise your deletion rights in respect of that resource, please contact us at support@gptbots.ai. Our staff will relay your deletion request to the third-party voice model service provider and assist you with the request to the extent technically feasible. We do not guarantee that a third-party voice model service provider outside our control will complete the deletion;

(4) Deletion of a Platform-Trained Voice is irreversible and cannot be undone. The voice quota used by a deleted Platform-Trained Voice will not be restored.

5.3 When you click to delete an Own Voice, we will unbind it from the GPTBots Platform and stop calling it, but we will not automatically delete the voice that you have trained, generated, or stored with the third-party voice model service provider. We will delete the SpeakerID and metadata within our control.

5.4 After your account on the GPTBots Platform is closed or your organization is dissolved, we will delete or deactivate the Custom Voices and related personal data under that account or organization and handle them in accordance with Sections 4 and 5 of these Terms. We may retain necessary copies or records only to the extent that retention is required or permitted by applicable law and there remains a specific purpose or legal basis. Any retained information will be subject to access restrictions and will not be used for other purposes, and will be deleted or de-identified once the basis for retention no longer applies.

6. Voice Quotas and Fees

6.1 The number of Platform-Trained Voices you may create is determined by the “voice quota” included in your paid subscription plan. Each successfully trained voice consumes one voice quota. No voice quota will be deducted for training failures caused by a GPTBots Platform or third-party voice model service provider failure, cancellation, duplicate requests, or circumstances not attributable to you. A deleted voice will not restore any voice quota already used. If you need additional voice quota, you may upgrade your subscription plan for an additional fee.

6.2 The number of retraining attempts permitted for each Platform-Trained Voice is subject to the training-attempt limit set by the third-party voice model service provider. Attempts that do not successfully generate a trained voice because of a GPTBots Platform or third-party voice model service provider failure, training failure, system retry, withdrawal/cancellation, or failure to meet the provider’s minimum technical requirements will not count toward that limit. Retraining will overwrite the previously trained Platform-Trained Voice, which cannot be restored after being overwritten. One voice quota supports the retention of only one successfully trained Platform-Trained Voice.

6.3 Each time you use a Platform-Trained Voice to submit a speech-synthesis request through the GPTBots Platform and call a third-party voice model service to generate synthesized speech (including previews), the applicable amount will be consumed. Third-party fees incurred when you use an Own Voice in Own Credential Mode to call a third-party voice model service and generate synthesized speech are settled directly between you and the third-party voice model service provider.

6.4 After your paid subscription plan expires or is downgraded, you will no longer be able to create new Platform-Trained Voices. However, you may continue to use Platform-Trained Voices created before the subscription plan expired to synthesize speech, and such synthesized speech will consume the corresponding GPTBots credits in your account.

7. Remedies for Breach

7.1 For security, abuse prevention, compliance with legal obligations, or handling valid complaints, we may conduct automated or manual reviews proportionate to the risks in relation to Voice Materials you upload to the GPTBots Platform, Own Voices you bind on the Platform, Synthesized Content we receive, and your use of the Feature. Our reviews will follow least-privilege, access-logging, and confidentiality controls. Any review does not constitute confirmation or a warranty that the foregoing Voice Materials, Own Voices, or Synthesized Content are lawful, properly owned, or suitable for use.

7.2 If we have reasonable grounds to believe that you have violated these Terms, that there is a material security or legal risk, or if we receive a facially credible rights complaint or valid legal requirement, we may take necessary and proportionate measures, including requiring you to provide an explanation or proof of rights within a reasonable period, restricting or disabling the voice concerned, or suspending the relevant Feature. In an emergency or where required by law, we may act without prior notice, but, where legally permitted, will notify you as soon as reasonably practicable and provide a channel for review. If you commit a serious, continuing, or irremediable breach, or a material security or legal risk arises, we may immediately discontinue the Service. When disclosing personal data to a rights claimant or competent authority, we will disclose it as required by law, through valid procedures, and only to the extent necessary, and will notify you where legally permitted.

7.3 To the fullest extent permitted by applicable law, you will indemnify us and our affiliates against third-party claims, reasonable legal fees, and related losses arising from your material breach of these Terms, your provision of Voice Materials or Own Voices without authorization, or your unlawful use of Synthesized Content. If any such third-party claim arises, we will notify you in writing of the claim and provide reasonable assistance. Any defense or settlement led by you must not prejudice our interests or admit liability on our behalf.

7.4 You will bear any losses you incur as a result of measures we take proportionately to the risks under this Section 7 of these Terms.

8. Disclaimer

8.1 The Feature depends on third-party voice model services. Voice-training results, similarity, availability, accuracy, and suitability of outputs may vary depending on the Voice Materials you upload, Own Voices you bind, other input, the third-party voice model, and the use case. Except as expressly required by the Main Agreement or applicable law, we do not guarantee any particular training result or suitability for a particular purpose; you must independently verify and legally review Synthesized Content before using it.

8.2 If a third-party voice model service is suspended, discontinued, materially changed, or restricted, we will notify you within a reasonable period after becoming aware of the matter and then suspend, replace, or terminate the affected Feature. To the extent permitted by applicable law, we will not be liable for your indirect losses arising from an interruption of a third-party voice model service that is outside our control and cannot be avoided despite reasonable mitigation.

8.3 Section 7 of the Main Agreement and these Terms must be read together. This section does not expand the scope of any disclaimer permitted under the Main Agreement and does not exclude any liability that cannot lawfully be excluded or limited. If the two documents conflict, these Terms shall prevail; however, applicable law prevails with respect to personal-data protection and mandatory legal obligations.

9. Updates and Effective Date

9.1 These Terms take effect when you accept them by checking the acceptance box or by another express method. We may update these Terms due to applicable law, service operations, or security needs, and will notify you in a reasonable manner of the updates, their effective date, and how to review the updated Terms. Non-material updates may take effect on the date specified in the notice. If an update materially and adversely affects your important rights or obligations, we will give you reasonable advance notice before it takes effect, and it will become effective for you only after you accept it by checking the box again or through another clear, express, and recordable affirmative action. If you do not accept such a materially adverse update, you may stop using the affected Service before the update takes effect. If applicable law requires an update to be implemented on a particular date or within a particular period, we may implement it to the extent and within the timeframe required by law and will notify you as early as legally permitted. If necessary to address an urgent security risk, we may first implement necessary and proportionate interim security measures and explain the measures and their impact as soon as practicable. Unless otherwise required by applicable law, updates apply prospectively only and do not retroactively alter processing completed or rights and obligations accrued before the update takes effect.

9.2 The governing law and dispute resolution provisions applicable to these Terms are set out in Section 8 of the Main Agreement. If they conflict with any mandatory provision of law, that mandatory provision will prevail.