Interactive listening evidence · Base & Medium

97 emotion and voice rankings

Choose a dimension to listen to the ten highest Base and Medium predictions.

The top ten examples on each page were selected from 1,000 Emolia clips using each model’s predictions. They illustrate what the heads rank highly; they do not measure agreement with human judgments.

40 EmoNet emotions

57 VoiceNet dimensions

Perceived voice age (AGEV)Arousal (AROU)Arousal shift (ARSH)Vocal attack (ATCK)Background noise (BKGN)Brightness (BRGT)Speech chunking (CHNK)Articulation clarity (CLRT)Cognitive load (COGL)Dynamic arc (DARC)Disfluency (DFLU)Emphasis (EMPH)Esthetics (ESTH)Content appropriateness (EXPL)Focus (FOCS)Fullness (FULL)Perceived gender (GEND)Harmonicity (HARM)Metallic character (METL)Pitch range (RANG)Recording quality (RCQL)Vocal register (REGS)Respiration (RESP)Roughness (ROUG)Chest resonance (R_CHST)Head resonance (R_HEAD)Mask resonance (R_MASK)Mixed resonance (R_MIXD)Nasal resonance (R_NASL)Oral resonance (R_ORAL)Throat resonance (R_THRT)Smoothness (SMTH)Stance (STNC)Speech structure (STRU)ASMR style (S_ASMR)Authoritative style (S_AUTH)Cartoonish style (S_CART)Casual style (S_CASU)Conversational style (S_CONV)Dramatic style (S_DRAM)Formal style (S_FORM)Monologue style (S_MONO)Narrator style (S_NARR)Newsreader style (S_NEWS)Playful style (S_PLAY)Ranting or angry style (S_RANT)Storytelling style (S_STRY)Teacher or didactic style (S_TECH)Whispered talking style (S_WHIS)Tempo (TEMP)Tension (TENS)Valence (VALN)Valence shift (VALS)Velocity flux (VFLX)Volatility (VOLT)Vulnerability (VULN)Warmth (WARM)