qwen voice demo
session
task
accent
gender
all
female
male
voice
preview
stop
delete clone
voice clone
audio file
name
clone voice
10 to 20 seconds of one clear voice, WAV MP3 or M4A, mono, under 10 MB
model
transport
websocket
webrtc
turn detection
vad
threshold
silence ms
idle timeout ms
prefix padding ms
replies automatically
can be interrupted
turn detection applies at start task
advanced
temperature
top_p
top_k
max_tokens
repetition
presence
seed
style
default
spoken
formal
show what you say
web search
start task
end task
apply settings & restart
not started
results
0
human turns
0
bot turns
0s
duration
0
barge-ins
–
reply latency
download audio
download transcript
transcript