~/talk

Enhancing XR Audio Realism via Multimodal Scene-Aware Acoustic Rendering

on YouTube 13m
Talkat UIST 2025 · 2025-09-28..10-01· [webxr]

speakers

name:Tianyu Xu
links:none on file
name:Jihan Li
links:none on file
name:Penghe Zu
links:none on file
name:Pranav Sahay
links:none on file
name:Maruchi Kim
links:none on file
name:Kwadwo Obeng-Marnu
links:none on file
name:Farley Miller
links:none on file
name:Xun Qian
links:none on file
name:Katrina Passarella-Ward
links:none on file
name:Mahitha Rachumalla
links:none on file
name:Rajeev Nongpiur
links:none on file
name:D Shin
links:none on file

resources

video:https://youtu.be/_b2SpmKSS9g
slides:none on file

summary

Samosa is an on-device hybrid pipeline that combines geometric ray-based shoebox estimation with learned acoustic priors to synthesize dynamic RIRs and convolve them with source audio in real time using roughly 3% of headset compute. It uses camera and depth feeds plus DeepLabV3+ for material masks refined via Segment Anything and a VLM, trains acoustic embeddings from ambisonic recordings across 25 rooms, and reports RT60/EDT parity with offline methods and a N=12 expert study showing better naturalness and externalization.

AI-written summary · about

chapters (7)

  1. 0:00Introduction to XR audio realism
  2. 2:54Challenge in XR audio rendering
  3. 3:55Simosa system overview
  4. 5:08Input and pipeline components
  5. 7:14RIR synthesis and applications
  6. 8:16Evaluation and future insights
  7. 10:05Q&A session

related talks

~/search
type to search 6,249 conferences, 138,987 speakers and 185,172 talks
~/help
cmdctrl k
search every conference, speaker and series
/
focus the filter search
h
home
c
conferences
s
speakers
e
series
t
light / dark — the status bar button does the same
?
this help
esc
close this, then clear every filter — or just the search, while you are typing in it

Ignored while you are typing. Pages without a filter rail have no search to focus.