Brazilian Portuguese Speech Data | Read and Spontaneous | 5 Regions | Consented for AI Training

Native Brazilian Portuguese speech recorded by adult contributors across the five regions of Brazil through our web app, on their own phones. Collection runs through Verakin, our contributor network in Brazil, recruited and supported in Portuguese. Speech styles - Read prompts: a shared core set read by every speaker and a split bank of sentences - Spontaneous monologues of 1 to 3 minutes on everyday topics Audio - WAV PCM 16-bit, mono, 48 kHz - Original file, with no conversion, normalization or trimming - Device echo cancellation, noise suppression and gain control requested off Metadata per clip - Speaker: gender, age band, region, state and self-reported accent - Device: model, operating system and browser - Audio measures: duration, loudness, peak, estimated SNR, clipping ratio and SHA-256 Transcripts - Spontaneous speech transcribed verbatim and reviewed by a native speaker - Read speech referenced to its prompt Rights and consent - Consent and an AI-training license recorded for each file, with terms version, acceptance time and text hash - The consent ledger ships with the data - Withdrawal of consent blocks future licensing and does not affect datasets already licensed - Processed under Brazil's LGPD Availability - Custom collection to your spec: region, age, gender, accent, prompts and domain - A general Brazilian Portuguese speech dataset is in construction for non-exclusive licensing - Evaluation sample available under NDA Delivery - Download link or S3-compatible