aboutsummaryrefslogtreecommitdiff
path: root/app/converter/config.py
diff options
context:
space:
mode:
Diffstat (limited to 'app/converter/config.py')
-rw-r--r--app/converter/config.py34
1 files changed, 22 insertions, 12 deletions
diff --git a/app/converter/config.py b/app/converter/config.py
index ae7189f..4694e80 100644
--- a/app/converter/config.py
+++ b/app/converter/config.py
@@ -25,21 +25,31 @@ BACKEND = "audiocpp"
# BACKEND 1: qwen-tts-demo (qwen) options #
###############################################################################
-# There are different API URLs for CustomVoice and Base models so you can run both at once
-QWEN_API_URL = "http://127.0.0.1:7860" # CustomVoice model
-CLONE_API_URL = "http://127.0.0.1:7861" # Base model
-
-# Remote (externally-run) server URLs. The hub probes these and offers a
-# "[remote]" backend entry when one answers with the expected backend, so an
-# externally-started server can be used alongside a locally-managed one.
-# Leave empty to disable remote probing for that backend. The defaults match
-# the local ports so an external server squatting the local port is found
-# without any configuration.
-QWEN_REMOTE_URL = "http://127.0.0.1:7860" # CustomVoice model
-CLONE_REMOTE_URL = "http://127.0.0.1:7861" # Base model
+# The qwen backend runs ONE demo server at a time, on this port. Which model
+# the server hosts is chosen per run on the Generate audiobooks screen and
+# persisted below (see QWEN_MODEL); switching models restarts the server.
+QWEN_API_URL = "http://127.0.0.1:7860" # single qwen-tts-demo server
+
+# Remote (externally-run) server URL. The hub probes it and offers a
+# "[remote]" backend entry when it answers with a known qwen-tts demo (any
+# of the three models), so an externally-started server can be used alongside
+# a locally-managed one. Leave empty to disable remote probing. The default
+# matches the local port so an external server squatting the local port is
+# found without any configuration.
+QWEN_REMOTE_URL = "http://127.0.0.1:7860"
+
+# Which model the managed demo server runs (one server hosts one model):
+# CustomVoice - built-in speakers (see SPEAKER)
+# Base - voice cloning from a reference .wav
+# VoiceDesign - voice described by an instruction
+# Chosen per run in the Generate-audiobooks form; edited here only as the
+# default for the next run.
+QWEN_MODEL = "CustomVoice"
# Custom voice options
SPEAKER = "Vivian" #Vivian, Serena, Uncle_Fu, Dylan, Eric, Ryan, Aiden, Ono_Anna, Sohee
+# Style/delivery instruction for CustomVoice runs; also the default design
+# instruction when a VoiceDesign run does not override it.
INSTRUCT = "Speak naturally and clearly, as if reading a dramatic book to an adult audience."
# Don't clone with transcription, only use x-vector-only cloning. Generally "worse"