Commercial training data from specialist providers. Listings show what each source advertises; licensing is confirmed with the provider before any agreement.
Real Sundanese, spoken the way it is actually lived.
This dataset provides textual summaries of real customer service calls, extracted and written by humans based on complete support interactions.
The complete operational email archive of a Ukrainian rail-freight logistics company — approximately 1.2 TB of correspondence and the documents attached to it, including…
Dataset containing unscripted Arabic speech from native speakers.
Audio dataset of scripted Arabic speech from native speakers.
Arabic audio dataset featuring native speakers reading scripted dialogues.
The provider is an audio production studio from Kragujevac, Serbia with over 15 years of expertise.
Multi-Modal Dataset — Noise, Mobility & POI Visitation This dataset combines three critical data streams essential for training multi-modal AI models and powering urban…
Noise Complaint Dataset — AI Training & Acoustic Source Detection Silencio offers the world’s largest noise complaint dataset, consisting of over 160,000 geolocated and…
This dataset contains audio recordings of daily life voice notes in India.
Finnish TTS dataset containing native speaker recordings of scripted man-machine dialogues.
Indonesian audio dataset with free talk sessions from native speakers.
1. Overview the provider has off-the-shelf 1 million videos of multimodal data, including multi-style video, lip syn, emotion video and etc.
Acapella and instrumental pairs is the clean ground truth for separation: 1,334 pairs, 2,668 WAV files, about 146 hours of audio, counted on 14 October 2026.
DAW sessions is the set nobody else sells: about 12,400 distinct project files from working electronic producers, counted in a census of every archive WMDM holds on 16…
Stems is the largest part of the the provider catalogue: 2,819 stems packs from finished electronic tracks, more than 63,000 stems, about 4,200 hours of audio, counted…
Norwegian-language conference calls involving 3–5 native speakers per call.
Norwegian-language TTS dataset with audio recordings of native speakers reading scripted dialogues.
Swedish TTS dataset with spoken sentences read by native speakers.
Audio dataset with unscripted Turkish speech.
Turkish audio dataset of scripted reading by native speakers.
Silencio’s Venue Noise-Level Dataset offers acoustic profiles of venues worldwide, collected from over 10 million hours of real dBA measurements.