c86e323d7a
Quality-first multi-pass audiobook generation from English manuscripts. Includes: - umbraaudio/ core package (pipeline, config, voices, web review UI) - scripts/ CLI utilities + model download script - tts_server/ remote TTS server - specs/ design docs, tests/, pyproject.toml + uv.lock - Qwen3-TTS git submodule (pinned 022e286) with local demo.py UI patch saved at scripts/qwen3_tts_demo_umbra_patch.patch - .env with runtime config/secrets Excluded (see .gitignore): models/ (downloaded via scripts/download_models.sh), output/, logs/, .venv, .umbra runtime data, and book projects (DealHunterShip_*), which are committed separately.
29 lines
965 B
Plaintext
29 lines
965 B
Plaintext
|
|
During utterance generation add modifiers for the local chapter/scene needs. Argumentative, soft, pleading, etc.
|
|
|
|
Question tone modifier.
|
|
|
|
Generate verbage to identify new speaker (first time) as narrator bit, or other speaker saying who they are.
|
|
|
|
Gene says, "Hello Princess."
|
|
or Robot says, "Welcome, Gene."
|
|
"Hello." - in Gene's voice.
|
|
"How are you today Princess?" - question attribute, Gene's voice.
|
|
"Good morning. I'm doing well." - Princess voice.
|
|
|
|
Minimize narrator speaking "X said, X says,"
|
|
|
|
***
|
|
Finding characters is simply finding names, nicknames, alias, references (the AI).
|
|
Speakers is someone with a quoted string.
|
|
|
|
Character can say a name and the speaker attribution should be normal.
|
|
|
|
"Hello Princess."
|
|
Kainda replies, "Good Morning"
|
|
Or
|
|
"Hello Princess."
|
|
"Good Morning" - spoken by Kainda's voice.
|
|
|
|
Need a prompt that will go through and check / find these and make sure it is the correct speaker. After / followup to correct, using context.
|
|
*** |