Find recurring voices
Group recurring voices across audio and video files, keep stable voice clusters, and show which files contain the same speaker
Uses a public CAM++ speaker model on speech windows from your selected files. Returns candidate voice groups and their time spans, keeps cluster embeddings in your account for later files, and shows a person's name only when a name was explicitly attached.
What it asks for
- files (Needed, list): One to twenty audio or video file identities from your Files
- max_seconds_per_file (Optional, number): Maximum seconds to inspect in each file (default 1800)
- minimum_similarity (Optional, number): Cosine similarity required to reuse a voice cluster (default 0.72)
- known_voices (Optional, list): Existing voice clusters from your account; filled automatically
- person_bindings (Optional, list): Names explicitly attached to voice clusters; filled automatically
Use it from your AI
Connect the AI you already use to Trillion once. Then ask in your own words; your AI finds this tool by what it does and runs it.
Say to your AI: “Use Trillion to group recurring voices across audio and video files, keep stable voice clusters, and show which files contain the same speaker”
What your AI calls: run_tool {"tool_id":"tool:find-recurring-voices"}