sij/sijapi
1
1
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

Auto-update: Thu Aug 8 14:30:29 PDT 2024

Browse source
This commit is contained in:
sanj 2024-08-08 14:30:29 -07:00
parent 6897e51fa0
commit 3d7ae743e6
3 changed files with 25 additions and 11 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

1
sijapi/__init__.py
View file

@ -29,6 +29,7 @@ News = Configuration.load('news', 'secrets')
Archivist = Configuration.load('archivist') Archivist = Configuration.load('archivist')
Scrape = Configuration.load('scrape', 'secrets', Dir) Scrape = Configuration.load('scrape', 'secrets', Dir)
Serve = Configuration.load('serve') Serve = Configuration.load('serve')
Tts = Configuration.load('tts')
# Directories & general paths # Directories & general paths
ROUTER_DIR = BASE_DIR / "routers" ROUTER_DIR = BASE_DIR / "routers"

24
sijapi/config/tts.yaml-example
View file

@ -1,10 +1,24 @@
default: xtts default: xtts
email: xtts email: xtts
webclip: 11L webclip: elevenlabs
rss: xtts rss: xtts
podcast_dir: '{{ DIR.HOME }}/Library/Mobile Documents/iCloud~co~supertop~castro/Documents/Sideloads'
xtts: xtts:
voice: joanne default: joanne
elevenlabs: elevenlabs:
voice: luna default: Victoria
api_key: '{{ SECRET.ELEVENLABS_API_KEY }}' voices:
alloy: "E3A1KVbKoWSIKSZwSUsW"
echo: "b42GBisbu9r5m5n6pHF7"
fable: "KAX2Y6tTs0oDWq7zZXW7"
onyx: "clQb8NxY08xZ6mX6wCPE"
nova: "6TayTBKLMOsghG7jYuMX"
shimmer: "E7soeOyjpmuZFurvoxZ2"
Luna: "6TayTBKLMOsghG7jYuMX"
Sangye: "E7soeOyjpmuZFurvoxZ2"
Herzog: "KAX2Y6tTs0oDWq7zZXW7"
Attenborough: "b42GBisbu9r5m5n6pHF7"
Victoria: "7UBkHqZOtFRLq6cSMQQg"
api_key: "{{ SECRET.ELEVENLABS_API_KEY }}"

11
sijapi/routers/tts.py
View file

@ -27,7 +27,7 @@ import tempfile
import random import random
import re import re
import os import os
from sijapi import L, Dir, DEFAULT_VOICE, TTS_SEGMENTS_DIR, VOICE_DIR, PODCAST_DIR, TTS_OUTPUT_DIR, ELEVENLABS_API_KEY from sijapi import L, Dir, Tts, DEFAULT_VOICE, TTS_SEGMENTS_DIR, VOICE_DIR, PODCAST_DIR, TTS_OUTPUT_DIR, ELEVENLABS_API_KEY
from sijapi.utilities import sanitize_filename from sijapi.utilities import sanitize_filename
### INITIALIZATIONS ### ### INITIALIZATIONS ###
@ -40,7 +40,6 @@ def err(text: str): logger.error(text)
def crit(text: str): logger.critical(text) def crit(text: str): logger.critical(text)
DEVICE = torch.device('cpu') DEVICE = torch.device('cpu')
MODEL_NAME = "tts_models/multilingual/multi-dataset/xtts_v2"
@tts.get("/tts/local_voices", response_model=List[str]) @tts.get("/tts/local_voices", response_model=List[str])
async def list_wav_files(): async def list_wav_files():
@ -211,7 +210,7 @@ async def determine_voice_id(voice_name: str) -> str:
"Victoria": "7UBkHqZOtFRLq6cSMQQg" "Victoria": "7UBkHqZOtFRLq6cSMQQg"
} }
if voice_name in hardcoded_voices: if voice_name in Tts.elevenlabs.voices:
voice_id = hardcoded_voices[voice_name] voice_id = hardcoded_voices[voice_name]
debug(f"Found voice ID - {voice_id}") debug(f"Found voice ID - {voice_id}")
return voice_id return voice_id
@ -232,7 +231,7 @@ async def determine_voice_id(voice_name: str) -> str:
except Exception as e: except Exception as e:
err(f"Error determining voice ID: {str(e)}") err(f"Error determining voice ID: {str(e)}")
# as a last fallback, rely on David Attenborough # as a last fallback, rely on David Attenborough; move this to tts.yaml
return "b42GBisbu9r5m5n6pHF7" return "b42GBisbu9r5m5n6pHF7"
@ -339,7 +338,7 @@ async def local_tts(
voice_file_path = await get_voice_file_path(voice, voice_file) voice_file_path = await get_voice_file_path(voice, voice_file)
# Initialize TTS model in a separate thread # Initialize TTS model in a separate thread
XTTS = await asyncio.to_thread(TTS, model_name=MODEL_NAME) XTTS = await asyncio.to_thread(TTS, model_name=Tts.xtts.model)
await asyncio.to_thread(XTTS.to, DEVICE) await asyncio.to_thread(XTTS.to, DEVICE)
segments = split_text(text_content) segments = split_text(text_content)
@ -397,7 +396,7 @@ async def stream_tts(text_content: str, speed: float, voice: str, voice_file) ->
async def generate_tts(text: str, speed: float, voice_file_path: str) -> str: async def generate_tts(text: str, speed: float, voice_file_path: str) -> str:
output_dir = tempfile.mktemp(suffix=".wav", dir=tempfile.gettempdir()) output_dir = tempfile.mktemp(suffix=".wav", dir=tempfile.gettempdir())
XTTS = TTS(model_name=MODEL_NAME).to(DEVICE) XTTS = TTS(model_name=Tts.xtts.model).to(DEVICE)
XTTS.tts_to_file(text=text, speed=speed, file_path=output_dir, speaker_wav=[voice_file_path], language="en") XTTS.tts_to_file(text=text, speed=speed, file_path=output_dir, speaker_wav=[voice_file_path], language="en")
return output_dir return output_dir