Fabio
c36510c21e
Some checks failed
Linux Build / Linux cpu (push) Has been cancelled
Linux Build / Linux cpu-clang (push) Has been cancelled
Linux Build / Linux vulkan (push) Has been cancelled
mac-build / macOS CPU (push) Has been cancelled
Nix Build / Nix (cpu) (push) Has been cancelled
Nix Build / Nix (vulkan) (push) Has been cancelled
windows-build / Windows CPU (push) Has been cancelled
79 lines
2.1 KiB
Python
79 lines
2.1 KiB
Python
#!/usr/bin/env python3
|
|
|
|
"""Sends one OpenAI-compatible speech request to a running audiocpp_server.
|
|
|
|
Model inference happens in C++ inside the server. This client uses only the
|
|
Python standard library.
|
|
"""
|
|
|
|
import json
|
|
import time
|
|
import urllib.error
|
|
import urllib.request
|
|
from pathlib import Path
|
|
|
|
|
|
REPO_ROOT = Path(__file__).resolve().parents[3]
|
|
|
|
SERVER_URL = "http://127.0.0.1:8080/v1/audio/speech"
|
|
|
|
TEXT = (
|
|
"আমি একটা টিটিএস মডেল, নাম OmniVoice। "
|
|
"কিভাবে আপনাকে সাহায্য করতে পারি?"
|
|
)
|
|
|
|
OUTPUT = (
|
|
REPO_ROOT
|
|
/ "tests/omnivoice/outputs/python_cpp_simple/python_output.wav"
|
|
)
|
|
|
|
payload = {
|
|
"model": "omnivoice",
|
|
"input": TEXT,
|
|
"language": "Bengali",
|
|
"seed": 42,
|
|
"num_inference_steps": 20,
|
|
"guidance_scale": 2.0,
|
|
"response_format": "wav",
|
|
"options": {
|
|
"speed": 1.2
|
|
}
|
|
}
|
|
|
|
request = urllib.request.Request(
|
|
SERVER_URL,
|
|
data=json.dumps(payload, ensure_ascii=False).encode("utf-8"),
|
|
headers={"Content-Type": "application/json"},
|
|
method="POST"
|
|
)
|
|
|
|
print("Sending request to audio.cpp server...")
|
|
|
|
start = time.perf_counter()
|
|
|
|
try:
|
|
with urllib.request.urlopen(request, timeout=600) as response:
|
|
audio_bytes = response.read()
|
|
except urllib.error.HTTPError as error:
|
|
body = error.read().decode("utf-8", errors="replace")
|
|
raise RuntimeError(
|
|
f"Server returned HTTP {error.code}: {body}"
|
|
) from error
|
|
except urllib.error.URLError as error:
|
|
raise RuntimeError(
|
|
"Could not connect to the audio.cpp server. "
|
|
"Run tests/omnivoice/python_cpp_simple/run_server.sh first."
|
|
) from error
|
|
|
|
elapsed = time.perf_counter() - start
|
|
|
|
if not audio_bytes.startswith(b"RIFF"):
|
|
preview = audio_bytes[:300].decode("utf-8", errors="replace")
|
|
raise RuntimeError(f"Expected WAV response, received: {preview}")
|
|
|
|
OUTPUT.parent.mkdir(parents=True, exist_ok=True)
|
|
OUTPUT.write_bytes(audio_bytes)
|
|
|
|
print(f"Python output: {OUTPUT}")
|
|
print(f"Response size: {len(audio_bytes)} bytes")
|
|
print(f"Request time: {elapsed:.3f} seconds")
|