Tuning Into Reality
Connecting to the world
ExperiencesConnecting to the world
A video, song, or advertisement sounds like you, but you never recorded those words. Your first task is to record where the audio appears and explain why it represents you. You can report suspected misuse without presenting an automated detector result as a definitive answer.
If it is being used to request money or passwords, warn affected people through a separate, trusted channel. If money has moved, contact the payment provider promptly and use the appropriate local fraud-reporting route. In the US, the FTC directs scam reports to ReportFraud.ftc.gov.
Voice misuse can feel particularly invasive because the recording appears to carry your identity, not just your work. You may be worried about customers believing an endorsement, an employer reusing old recordings, or listeners hearing you perform material you would never choose. Those situations need different explanations even when they all involve a suspected clone.
There may be a published audio clip, a fake account distributing it, and an underlying voice model capable of generating more material. Make a separate entry for each thing you have actually found. Removing one video is a useful outcome, but it does not establish that a model has been disabled or that other copies have disappeared.
If you only know about a video, begin there. You do not need to identify the exact generation software to describe an unauthorized representation of your voice. If you have a public voice-library entry or a message identifying a provider, preserve that lead and investigate it separately.
Record the direct URL, account handle, discovery date and time zone, exact timestamps, and what the voice says. Capture the caption, your name or image, product endorsement, payment request, or other context. Keep a permitted copy where available; if you only have a screen recording, identify it as such. Do not assume a screenshot preserves the audio.
Keep examples of authentic speech or singing that you are entitled to share. Before uploading them to another service, check its retention, sharing, training-use, and deletion terms. Keep your private investigation separate from a public post that could spread the imitation further.
Separate “this sounds like me” from “the uploader says it is me” and “the creator admitted using my recording.” Note pronunciation, characteristic phrasing, and the circumstances in which people identified you. Also record plausible alternatives, including a human impression or edited authentic audio.
An AI-generation detector and a comparison of speaker identity answer different questions. Neither result should be converted into a legal finding. NIST discusses the limitations and evaluation of synthetic-content detection, including the consequences of false positives.
Choose authentic examples whose origin you can explain: an episode you recorded, a dated studio session, or a video from your official channel. Identify the speaker, date, relevant timestamps, and any processing you know was applied. Keep a few representative examples instead of sending an unorganized archive.
Where possible, include context comparable to the disputed audio. A singing performance and a spoken advertisement may present different comparison difficulties. Note language, accent, background music, recording quality, and whether you are comparing normal speech with an exaggerated character performance. These details help a reviewer understand the material; they are not a checklist that guarantees identification.
Imagine you narrate educational videos. A new channel publishes a product promotion using a voice your regular audience identifies as yours. “That sounds like me” is the initial concern. A more useful packet identifies the promotion’s exact timestamps, your authentic narration references, your name in the caption, and messages from viewers who thought you endorsed the product.
That packet does not require you to claim that a particular app generated the audio. It explains the representation, why people connect it to you, and what you did not authorize. If you later receive an admission or identify a voice-library listing, add it as new evidence rather than rewriting your earlier suspicion as something you knew all along.
If an employer, client, studio, or former collaborator had recordings, collect the agreement and messages about permitted use. Record what you authorized, for which project, for how long, and whether synthetic generation was discussed. Owning a recording and having permission to create new speech are questions to examine separately; get local advice if the contract is disputed.
Start with the original project. What did you supply, and what does the agreement say about reuse, synthetic speech, additional campaigns, territory, and duration? Keep the signed version, amendments, and relevant emails together. If the other party says a clause authorizes the use, ask them to identify that clause and explain how it applies to this specific output.
You can state the observed use and ask for an explanation without deciding the contract dispute yourself. For example: “I supplied narration for the training video dated [date]. I have now found newly generated speech in [advertisement]. Please identify the authorization you rely on for this use and explain whether a voice model exists.” A representative or lawyer may be useful when a commercial relationship makes direct discussion difficult.
If the audio or model is connected to ElevenLabs, use its Unauthorized Use of Voice report. Its policy addresses cloning and impersonation without consent or other valid authorization. Include the distribution URLs and any model or voice identifier you have. Do not name a provider as responsible merely because the audio sounds synthetic.
Keep a simple log with one row per URL, report number, decision, and reappearance. Link related cases when the platform offers a way to do so. This is more useful than repeatedly submitting the same report without the requested detail. If a new upload changes the account or title, preserve its new identifiers.
When a provider is involved, ask what its response covers: the identified output, the voice-library listing, the account, or another specific object. Do not infer a broader action from a generic statement that the complaint was reviewed. Keep the reply and any stated limits.
For audience communication, use a short statement through an established channel: identify the content you did not record or endorse and tell people where authentic updates come from. If fraudulent requests are involved, provide a reliable verification method. You can correct the impression without posting additional copies of the imitation.
Commercial exploitation, a disputed contract, repeated identity misuse, and financial fraud can require different help. Collect the harm as well as the audio: customer confusion, campaign details, lost work you can document, or payment requests. Keep estimated losses distinct from confirmed transactions.
Use the country guidance to find the relevant privacy authority, fraud route, or local IP guidance. Ask a qualified adviser what remedy is available for your circumstances and what evidence is still needed. Avoid assuming that every voice dispute is a copyright dispute or that a platform decision settles local rights.
Use a factual summary: “At [URL], from [time] to [time], the audio represents me saying [description]. I did not record or authorize this use. These authentic references and contextual details identify me: [details]. Please review under the relevant synthetic-likeness or impersonation policy.” Adapt it to what is true, and complete the platform’s required fields.
Save the case number and decision. If refused, address the reason in the available review process. For privacy, contract, or financial remedies, use the country-specific help directory; a platform report and a legal claim serve different purposes.