Angry · Set 1
Reference
Generated music
piano, violin string, and drum, angry
piano, violin string, and drum, valence -0.75, arousal +0.25.
piano, violin string, and drum + speech tokens
SEAM · ICASSP 2027
Given a spoken utterance and an emotion-free instrument prompt, generate background music whose valence–arousal follows the speech — without putting mood words in the text.
Paper companion
Play the speech, then the three music clips. All systems share the same SA3 backbone. Ours uses speech tokens and an emotion-free prompt; the others stay text-only.
Reference
Generated music
piano, violin string, and drum, angry
piano, violin string, and drum, valence -0.75, arousal +0.25.
piano, violin string, and drum + speech tokens
Reference
Generated music
piano, violin string, and drum, angry
piano, violin string, and drum, valence -0.75, arousal +0.50.
piano, violin string, and drum + speech tokens
Reference
Generated music
piano, violin string, and drum, happy
piano, violin string, and drum, valence +0.50, arousal 0.00.
piano, violin string, and drum + speech tokens
Reference
Generated music
piano, violin string, and drum, happy
piano, violin string, and drum, valence +0.50, arousal +0.17.
piano, violin string, and drum + speech tokens
Reference
Generated music
piano, violin string, and drum, sad
piano, violin string, and drum, valence -0.50, arousal -0.25.
piano, violin string, and drum + speech tokens
Reference
Generated music
piano, violin string, and drum, sad
piano, violin string, and drum, valence -0.50, arousal +0.25.
piano, violin string, and drum + speech tokens
Reference
Generated music
piano, violin string, and drum, excited
piano, violin string, and drum, valence +0.25, arousal 0.00.
piano, violin string, and drum + speech tokens
Reference
Generated music
piano, violin string, and drum, excited
piano, violin string, and drum, valence +0.75, arousal +0.50.
piano, violin string, and drum + speech tokens
Task 2
Same spoken utterance, five α values. Ours steers the speech residual on an emotion-free prompt; SA3 writes the mix into the text and never sees the speech.
p with (1−α) of neutral and α of {emotion}.Reference
Ours
piano, violin string, and drum + speech tokens
piano, violin string, and drum + speech tokens
piano, violin string, and drum + speech tokens
piano, violin string, and drum + speech tokens
piano, violin string, and drum + speech tokens
SA3 lexical blend
piano, violin string, and drum with 1.00 of neutral and 0.00 of angry.
piano, violin string, and drum with 0.75 of neutral and 0.25 of angry.
piano, violin string, and drum with 0.50 of neutral and 0.50 of angry.
piano, violin string, and drum with 0.25 of neutral and 0.75 of angry.
piano, violin string, and drum with 0.00 of neutral and 1.00 of angry.
Reference
Ours
piano, violin string, and drum + speech tokens
piano, violin string, and drum + speech tokens
piano, violin string, and drum + speech tokens
piano, violin string, and drum + speech tokens
piano, violin string, and drum + speech tokens
SA3 lexical blend
piano, violin string, and drum with 1.00 of neutral and 0.00 of happy.
piano, violin string, and drum with 0.75 of neutral and 0.25 of happy.
piano, violin string, and drum with 0.50 of neutral and 0.50 of happy.
piano, violin string, and drum with 0.25 of neutral and 0.75 of happy.
piano, violin string, and drum with 0.00 of neutral and 1.00 of happy.
Reference
Ours
piano, violin string, and drum + speech tokens
piano, violin string, and drum + speech tokens
piano, violin string, and drum + speech tokens
piano, violin string, and drum + speech tokens
piano, violin string, and drum + speech tokens
SA3 lexical blend
piano, violin string, and drum with 1.00 of neutral and 0.00 of sad.
piano, violin string, and drum with 0.75 of neutral and 0.25 of sad.
piano, violin string, and drum with 0.50 of neutral and 0.50 of sad.
piano, violin string, and drum with 0.25 of neutral and 0.75 of sad.
piano, violin string, and drum with 0.00 of neutral and 1.00 of sad.
Reference
Ours
piano, violin string, and drum + speech tokens
piano, violin string, and drum + speech tokens
piano, violin string, and drum + speech tokens
piano, violin string, and drum + speech tokens
piano, violin string, and drum + speech tokens
SA3 lexical blend
piano, violin string, and drum with 1.00 of neutral and 0.00 of excited.
piano, violin string, and drum with 0.75 of neutral and 0.25 of excited.
piano, violin string, and drum with 0.50 of neutral and 0.50 of excited.
piano, violin string, and drum with 0.25 of neutral and 0.75 of excited.
piano, violin string, and drum with 0.00 of neutral and 1.00 of excited.