Create tts_handler.py
Browse files- tts_handler.py +29 -0
tts_handler.py
ADDED
|
@@ -0,0 +1,29 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# rentbot/tts_handler.py
|
| 2 |
+
import os
|
| 3 |
+
from elevenlabs.client import AsyncElevenLabs
|
| 4 |
+
from elevenlabs import stream
|
| 5 |
+
|
| 6 |
+
client = AsyncElevenLabs(api_key=os.getenv("ELEVENLABS_API_KEY"))
|
| 7 |
+
|
| 8 |
+
async def text_to_speech_stream(text_iterator):
|
| 9 |
+
"""
|
| 10 |
+
Streams text to ElevenLabs and yields audio chunks.
|
| 11 |
+
Twilio requires audio in 8-bit μ-law format.
|
| 12 |
+
"""
|
| 13 |
+
voice_id = os.getenv("ELEVENLABS_VOICE_ID", "21m00Tcm4TlvDq8ikWAM")
|
| 14 |
+
|
| 15 |
+
try:
|
| 16 |
+
audio_stream = await client.generate(
|
| 17 |
+
model="eleven_turbo_v2",
|
| 18 |
+
voice=voice_id,
|
| 19 |
+
text=text_iterator,
|
| 20 |
+
stream=True,
|
| 21 |
+
output_format="ulaw_8000", # Critical for Twilio
|
| 22 |
+
)
|
| 23 |
+
|
| 24 |
+
# Stream the audio chunks
|
| 25 |
+
async for chunk in audio_stream:
|
| 26 |
+
yield chunk
|
| 27 |
+
|
| 28 |
+
except Exception as e:
|
| 29 |
+
print(f"Error in TTS stream: {e}")
|