Datasets
Audio
Multi-Speaker Podcasts
Recorded as ongoing multi-episode shows rather than one-off sessions, multi-speaker podcasts capture extended dialogue between hosts, guests, and panels across the full range of contemporary podcast formats. The form covers interview shows, co-hosted banter, news and commentary panels, narrative storytelling with multiple voices, and roundtable discussions.
Hours312.3K+
Genres50+
Speakers2-8
Training Use Cases
✓Speaker diarization and turn-taking
✓Long-form speech recognition and transcription
✓Voice cloning across diverse speakers
✓Conversational AI training on natural multi-speaker dialogue
Key Highlights
✓50+ podcast genres covered including interview, comedy, news, true crime, business, and sports
✓2-to-8 speaker session sizes spanning solo hosts, co-hosts, panels, and rotating guests
✓Natural turn-taking, overlap, laughter, and crosstalk preserved as recorded
✓Episode-length recordings rather than highlight clips, capturing full conversational arcs
Metadata Fields
durationLength of episode in HH:MM:SS
sample_rateAudio sample rate (e.g., 44.1kHz, 48kHz)
channelsmono | stereo
languagePrimary spoken language (ISO 639-1 code)
primary_categoryDominant content category assigned to a recording
podcast_genreinterview | comedy | news | true_crime | business | science | politics | sports
speaker_countNumber of speakers in the episode
has_remote_speakersWhether the episode includes remote-recorded participants (boolean)
episode_formatfull_episode | segment