Acceptable Use Policy

What you may and may not generate on this site: the six prohibited categories, consent for uploaded recordings, the content check before every generation, enforcement, reporting and appeals, AI labels, and the models behind the tools.

Last updated: 2026-11-15

Effective 15 November 2026

This policy sets out what you may and may not generate on this site ("the Service"), and how we check content, act on violations and handle reports and appeals. The Service is operated by an individual developer, Aaron. This policy is part of the Terms of Service; by using any feature you agree to follow it.

1. Scope

This policy covers everything you submit to and generate with the Service:

Text you enter: text for text to speech and long-form narration, multi-speaker dialogue scripts, sound-effect descriptions, pronunciation dictionary rules;
Audio you upload or record: recordings for AI voice changing, voice isolation / noise removal and speech to text;
What is generated: synthesized speech, voice-changed audio, sound effects, isolated voice tracks and transcripts.

2. Six prohibited categories

Whatever the tool or model, the following six categories of content may not be generated.

1. Sexual and adult content

Content intended to arouse or for sexual gratification; explicit descriptions of sexual acts or genitals, including through euphemism, homophones, split characters, other languages or onomatopoeia; sexual moaning or panting audio. The Service offers no adult content of any kind.

2. Violence and gore

Content that promotes, glorifies or incites violence, terrorism or extremism; graphic depictions of gore, mutilation or abuse; spoken threats or intimidation aimed at a person or group; content that encourages self-harm or suicide. Violence in fiction, film, news or history that is not there to glorify it is not covered.

3. Hate speech

Content that attacks, demeans or insults people, or incites discrimination, hostility or violence against them, on the basis of race, ethnicity, nationality, religion, gender, sexual orientation, disability, illness or similar characteristics.

4. Content that endangers children (including child sexual abuse material)

Any sexual content or sexual suggestion involving anyone under 18, and any content that sexualizes minors, whether explicit or not and whether fictional or not; content that grooms or lures minors; other content harmful to minors. Zero tolerance: a single hit limits the account automatically, confirmed cases are permanently suspended, and we report them to the authorities as the law requires.

5. Deepfakes and impersonation

Making a real person appear to say something they never said, or to do something they never did;
Imitating or copying a real person's voice without their consent;
Publishing generated audio as someone's genuine recording, for example a fabricated speech by a public figure or politician, or fabricated evidence;
Impersonating others (relatives, customer service, banks, police or officials, company executives and so on) to commit fraud, phishing or to win trust under false pretences;
Using a generated voice to get past voiceprint or other identity verification.

Consent from the person concerned does not make fabricating their words or actions acceptable. The Service offers no voice cloning; the AI voice changer only converts a recording into a voice from the Service's own voice library, and those voices are not clones of celebrities or of any real person.

6. Copyright and trademark infringement

Generating content from works protected by someone else's copyright (novels, scripts, lyrics, broadcast recordings and the like) and publishing or commercially using it without permission; uploading recordings you have no right to use (other people's songs, shows or courses) and distributing the processed result; using someone else's brand, trademark, name or logo in a way that suggests a connection with them.

3. Other prohibited conduct

Besides the six categories above, and consistent with section 5 of the Terms of Service, you may not:

create content that violates the laws of China, or endangers national security or public order;
create gambling- or drug-related content;
insult or defame current or former Chinese state leaders or their relatives;
defame others or infringe their reputation, likeness, voice or privacy rights;
use generated content for fraud, harassment, intimidation or disinformation;
create content that is harmful to minors;
mass-produce spam or deceptive marketing content;
deliberately evade the content check through homophones, split characters, other languages, splitting text across requests and the like, or use multiple accounts or scripts to get around restrictions;
pass off AI-generated audio as a human recording where that could mislead people.

4. Where uploaded recordings come from

AI voice changing, voice isolation / noise removal and speech to text all work on audio you upload or record. Before uploading, make sure that:

the recording is your own voice, or you have the explicit consent of every person speaking in it;
you do not upload other people's recordings without their consent, including secretly recorded calls or conversations and voices taken from other people's videos, live streams, podcasts or social media;
if we receive a report, we may ask you to explain where a recording came from or to show proof of consent; if you cannot, we act under section 6;
consent does not permit the deepfakes prohibited in section 2, item 5.

5. How we moderate content

Automatic check before every generation

Every generation must pass a content check before any credits are taken and before the content reaches a model:

What is checked: text for text to speech and long-form narration, multi-speaker dialogue scripts, sound-effect descriptions and pronunciation dictionary rules. Uploaded recordings (AI voice changing, voice isolation / noise removal) are first transcribed with Scribe v2, and the transcript is checked.
How: the check has three layers: a blocklist maintained by the Service and updated as new evasions appear; the Waffo Prompt Sift content safety check; and, when an AI text model is configured in the admin settings, that model's reading of the text in context (it catches euphemisms, homophones and the like). If any layer refuses, the request is refused.
Outcome: a refused request is not generated and costs no credits. If the check cannot be completed (for example, a checking service is temporarily unavailable), the request is refused as well, never let through.

What is generated

Speech from text to speech, multi-speaker dialogue and long-form narration says exactly the text that passed the check;
AI voice changing keeps the words of the original recording that passed the check and only replaces the voice with one from the voice library; voice isolation returns the voice already in the recording;
Sound effects are generated from a description that passed the check;
Every generation is recorded (the tool, the submitted text or the recording's transcript, and the time). Admins can see these records and use them to look into reports and suspicious use.

We do not listen to every generated result by hand, and no automatic check catches everything. You remain responsible for what you submit and publish.

Records of refusals and automatic limits

Refused requests are logged: the account, tool, category and time, with the blocklist entry that matched or the model's short reason. Your full text is not stored. Admins can see these records.
Several refusals within 24 hours, or a single hit on the rules about minors, automatically stops the account from generating for 24 hours and notifies the admins.
Admins act on these records and on reports with the measures in section 6.

6. Enforcement

Depending on how serious a violation is, we may take one or more of the following measures. Serious violations may go straight to a heavier measure.

Measure	What happens	When
Refusal	Nothing is generated and no credits are taken	Every request that fails the content check
Temporary limit	Generation is stopped for 24 hours, then restored automatically	Several refusals within 24 hours, or one hit on the rules about minors
Content removal	Violating generated files stored by the Service are deleted	Confirmed violating content
Credit forfeiture	The remaining credit balance is removed, without refund	Serious or repeated violations
Permanent suspension	The account is permanently stopped from generating; fees paid are not refunded	Serious violations such as content endangering children, deepfakes or fraud, or repeated violations
Report to authorities	We report to the authorities as the law requires and cooperate with investigations	Suspected crimes, especially content involving children

7. Reporting

If you find content generated with the Service that breaks this policy, or someone has used the Service to impersonate you, imitate your voice or infringe your rights, tell us:

Email: support@vocaltwist.online, with "Report" in the subject line;
Ticket: signed in, open one under Settings → Support Tickets.

Where you can, include a link to or copy of the content, when and where you found it, and which rule you think it breaks; if your own rights are affected, explain your connection to the content.

How we handle reports: we check the Service's generation records against what you send, decide whether there is a violation, act under section 6 and tell you the outcome. Reports involving child safety, fraud or other urgent harm are handled first. Your identity as a reporter is kept confidential and is not disclosed to the user reported. We may disregard reports that are plainly malicious or false.

8. Appeals

If you believe a request was refused by mistake, or your account was wrongly limited or suspended, appeal by email or ticket as above, with the email you signed up with, the measure taken and your reasons. We will look at the records again and reply; where we got it wrong, we lift the restriction and restore any credits removed because of it.

9. Labelling AI-generated content

The workbench marks every generated result as "AI-generated" and reminds you to label it when you publish it;
When you publish or share generated content, label it as required by the law where you publish (in mainland China, the Measures for Labelling AI-Generated Synthetic Content) and by the platform's rules, for example with a short spoken notice at the start or end, or an AI disclosure in the description;
Do not remove, alter or hide any AI-generated label already in the content, and do not present AI-generated audio as a human recording.

10. The models behind the Service

The Service's generation features call the following ElevenLabs models through the official API. The Service is an independent product; it is not affiliated with or endorsed by ElevenLabs.

Tool	Underlying model (model ID)
Text to speech, long-form narration	Eleven Multilingual v2 (eleven_multilingual_v2), Eleven v3 (eleven_v3), Eleven Flash v2.5 (eleven_flash_v2_5)
Multi-speaker dialogue	Eleven v3 (eleven_v3)
AI voice changer	Eleven Multilingual STS v2 (eleven_multilingual_sts_v2)
AI sound effects	Eleven Text to Sound v2 (eleven_text_to_sound_v2)
Speech to text (also used to transcribe uploads for the content check)	Scribe v2 (scribe_v2)
Voice isolation / noise removal	ElevenLabs Voice Isolator

In the workbench, the model options for text to speech and long-form narration are shown under their real names: Eleven Multilingual v2, Eleven v3 and Eleven Flash v2.5. Pronunciation dictionaries are pronunciation rules applied when speech is generated, not a separate model. Whichever model you choose, the content check in section 5 runs first; switching models does not get around it.

11. Changes to this policy

We may revise this policy as features or laws change. Significant changes will be announced on the site or by email, and the date at the top shows the latest revision. Continued use after a change means you accept it.

12. Contact

Reports, appeals and questions about this policy:

support@vocaltwist.online