Where your recordings go

AI, yes. The first draft of the transcript comes from speech recognition with Whisper (run locally through whisper.cpp) and speaker separation with pyannote, both on our workstations in Brazil. Producing the transcript doesn’t send your audio to an outside service. A native speaker then reviews the draft, listens again wherever it reads wrong, and corrects it. For translation and the steps after it, we list the tools before the project starts, including any cloud service.

Files you send through our upload link wait on our web server in the United States until we move them to our workstations in Brazil. Your browser encrypts them before they leave your computer, the key that opens them isn’t on the server, and we delete the server copy once they’re moved, within 30 days of upload at the latest. Speech recognition and speaker separation run on our workstations in Brazil. Where the files sit during review, translation, and delivery, we state in writing for your project before you send anything.

By default, only Rivera Brazil staff assigned to your project, who need access to do the work, handle its files. We don’t give an outside specialist, such as a consultant with a graduate degree in your field, access to project files unless you approve that specialist in writing in advance. An approved specialist is bound by written confidentiality, data protection, and security obligations at least as protective as ours and, where applicable, is appointed as a subprocessor under the data processing agreement. Any service that stores or processes project files, such as the host of the upload link or a translation tool, is named in writing before the project starts and, where a data processing agreement applies, needs your authorization as a subprocessor.

Through a private upload link we set up for your project. Please don’t send recordings by email or through our contact form. For recordings from the EU or the EEA, no standard contractual clauses are needed to send them to us when they go to storage in Brazil or the EEA, because the European Commission has found that Brazil ensures an adequate level of protection (Commission Implementing Decision (EU) 2026/179); we name the upload service and its location in writing before you send anything; for recordings from the UK or Switzerland, tell us which transfer mechanism your institution requires before you send any audio.

Yes, after reviewing them with you. An NDA is in place before we receive any file, and for projects under the GDPR, a data processing agreement in which we act as your processor. We don’t accept health information protected by HIPAA unless the project has been approved for it in writing and a Business Associate Agreement is signed before any file is sent.

Yes, on request. Names, places, and other identifiers are replaced with codes you define, and the codes match in the Portuguese and the English.

As long as your protocol says. When that period ends, we delete the files and confirm the deletion by email.

We’re a Brazilian company, Agência Rivera de Consultoria em Comunicação, Tradução, Edição e Publicidade Ltda. (CNPJ 42.968.216/0001-95), subject to the LGPD, Brazil’s general data protection law.

Yes. We can run a pilot on a short excerpt of one of your recordings.

Each delivery comes with a short report listing the files processed, the conventions used, and who reviewed them.