from fastapi import FastAPI, File, UploadFile, Form
from fastapi.responses import HTMLResponse, FileResponse
import whisper
import torch
import os
import srt
import logging
from datetime import timedelta
from time import time
from fpdf import FPDF
import uvicorn
logging.basicConfig(filename="speech_transcriber.log", level=logging.INFO,
format="%(asctime)s [%(levelname)s] %(message)s")
logger = logging.getLogger(__name__)
device = "cuda" if torch.cuda.is_available() else "cpu"
app = FastAPI()
@app.get("/", response_class=HTMLResponse)
def home():
return f"""
Trascrizione Audio/Video
🎤 Trascrizione Audio/Video
"""
@app.post("/upload/")
async def upload_audio(
file: UploadFile = File(...),
format: str = Form(...),
model_name: str = Form("base")
):
os.makedirs("temp", exist_ok=True)
base = os.path.splitext(file.filename)[0]
path = os.path.join("temp", file.filename)
with open(path, "wb") as f:
f.write(await file.read())
model = whisper.load_model(model_name, device=device)
result = model.transcribe(path)
subtitles = [
srt.Subtitle(i+1, timedelta(seconds=s['start']), timedelta(seconds=s['end']), s['text'].upper())
for i, s in enumerate(result["segments"])
]
def to_tc(t, fps=25):
sec = t.total_seconds()
h = int(sec // 3600)
m = int((sec % 3600) // 60)
s = int(sec % 60)
f = int((sec % 1) * fps)
return f"{h:02}:{m:02}:{s:02}:{f:02}"
ext = "txt" if format == "avid" else format
output_path = os.path.join("temp", f"{base}.{ext}")
if format == "avid":
with open(output_path, "w") as out:
out.write("@ This file written with MonniS Avid Caption plugin, version 1\n\n\n")
for s in subtitles:
out.write(f"{to_tc(s.start)} {to_tc(s.end)}\n{s.content}\n\n")
out.write("")
elif format == "xml":
with open(output_path, "w") as out:
out.write('\n\n')
for s in subtitles:
out.write(f' {s.content}\n')
out.write("")
elif format == "pdf":
pdf = FPDF()
pdf.set_font("Arial", size=10)
def add_header(tc_str):
pdf.add_page()
try:
if os.path.exists("logosermi.png"):
pdf.image("logosermi.png", x=(pdf.w - 50) / 2, y=10, w=50)
pdf.set_y(40)
else:
pdf.set_y(20)
except Exception as e:
logger.warning(f"Logo non inserito: {e}")
pdf.set_y(20)
pdf.set_x(0)
pdf.cell(w=0, h=10, txt=f"Da: {tc_str}", align="C", ln=True)
pdf.set_y(pdf.get_y() + 10)
first_in_page = True
for i, s in enumerate(subtitles):
if first_in_page or pdf.get_y() > 270:
tc = s.start.total_seconds()
tc_str = f"{int(tc//3600):02}:{int((tc%3600)//60):02}:{int(tc%60):02}:{int((tc%1)*25):02}"
add_header(tc_str)
first_in_page = False
line = s.content.encode("latin-1", "replace").decode("latin-1")
pdf.multi_cell(0, 10, line)
if pdf.get_y() > 270:
first_in_page = True
pdf.output(output_path, "F")
else:
return HTMLResponse("❌ Formato non supportato", status_code=400)
return FileResponse(output_path, media_type="application/octet-stream", filename=os.path.basename(output_path))
if __name__ == "__main__":
uvicorn.run(app, host="0.0.0.0", port=10000)