Initial commit: SmallWhisper - web app per trascrizione audio con faster-whisper

app.py

@@ -0,0 +1,126 @@
+import os
+import uuid
+import threading
+from flask import Flask, request, render_template, send_file, url_for
+from faster_whisper import WhisperModel
+
+app = Flask(__name__)
+app.config['MAX_CONTENT_LENGTH'] = 100 * 1024 * 1024  # 100 MB limit
+app.secret_key = os.urandom(24)
+
+# ── Whisper model ───────────────────────────────────────────────
+# Modello small (1.5 GB) — ottimo per italiano, veloce su CPU.
+# Cambia in "base" o "tiny" se vuoi più velocità, "medium" per più accuratezza.
+WHISPER_MODEL = "small"
+
+print(f"[INFO] Caricamento modello Whisper '{WHISPER_MODEL}'...")
+whisper = WhisperModel(WHISPER_MODEL, device="cpu", compute_type="int8")
+print("[INFO] Modello caricato!")
+
+# ── Job tracking (in memoria) ───────────────────────────────────
+jobs = {}
+
+# ── Routes ──────────────────────────────────────────────────────
+
+@app.route('/')
+def index():
+    return render_template('index.html')
+
+
+@app.route('/upload', methods=['POST'])
+def upload():
+    if 'audio' not in request.files:
+        return {'error': 'Nessun file audio'}, 400
+    file = request.files['audio']
+    if file.filename == '':
+        return {'error': 'Nome file vuoto'}, 400
+
+    job_id = str(uuid.uuid4())
+    os.makedirs('tmp', exist_ok=True)
+    tmp_path = os.path.join('tmp', f'{job_id}.m4a')
+    file.save(tmp_path)
+
+    jobs[job_id] = {
+        'status': 'running',
+        'progress': 0,
+        'message': 'Caricato',
+        'download': None,
+    }
+
+    threading.Thread(target=job_worker, args=(job_id, tmp_path), daemon=True).start()
+
+    return {'statusUrl': url_for('status', job_id=job_id)}
+
+
+@app.route('/status/<job_id>')
+def status(job_id):
+    job = jobs.get(job_id)
+    if not job:
+        return {'error': 'Job non trovato'}, 404
+
+    # Mentre whisper lavora, mostriamo "Trascrizione in corso..." con progress 50
+    if job.get('status') == 'running' and job.get('progress', 0) < 100:
+        return {
+            'status': 'running',
+            'progress': 50,
+            'message': 'Trascrizione in corso...',
+            'download': None,
+        }
+
+    return {
+        'status': job.get('status'),
+        'progress': job.get('progress', 100),
+        'message': job.get('message'),
+        'download': job.get('download'),
+    }
+
+
+@app.route('/download/<job_id>.txt')
+def download(job_id):
+    txt_path = os.path.join('tmp', f'trascrizione_{job_id}.txt')
+    if not os.path.exists(txt_path):
+        return {'error': 'File non trovato'}, 404
+    return send_file(
+        txt_path,
+        as_attachment=True,
+        download_name='trascrizione.txt',
+        mimetype='text/plain',
+    )
+
+
+# ── Worker (thread background) ──────────────────────────────────
+
+def job_worker(job_id, tmp_path):
+    try:
+        jobs[job_id]['message'] = 'Trascrizione in corso...'
+
+        # Whisper trascrive direttamente dal file (supporta .m4a via ffmpeg)
+        segments, info = whisper.transcribe(tmp_path, language="it", beam_size=5)
+
+        transcript_parts = []
+        for segment in segments:
+            transcript_parts.append(segment.text.strip())
+
+        transcription = "\n".join(transcript_parts)
+
+        txt_path = os.path.join('tmp', f'trascrizione_{job_id}.txt')
+        with open(txt_path, 'w', encoding='utf-8') as f:
+            f.write(transcription)
+
+        jobs[job_id]['download'] = f'/download/{job_id}.txt'
+        jobs[job_id]['message'] = 'Completato'
+        jobs[job_id]['progress'] = 100
+        jobs[job_id]['status'] = 'done'
+
+    except Exception as e:
+        jobs[job_id]['status'] = 'error'
+        jobs[job_id]['message'] = str(e)
+    finally:
+        try:
+            os.remove(tmp_path)
+        except Exception:
+            pass
+
+
+if __name__ == '__main__':
+    app.run(host='0.0.0.0', port=5000, debug=True)