Spaces:
Running on Zero
Running on Zero
Transcode uploads to WAV before analysis
Browse files
app.py
CHANGED
|
@@ -6,6 +6,8 @@ Zugangsschlüssel (Secret APP_PASS); erst nach der Prüfung wird GPU-Zeit
|
|
| 6 |
verbraucht.
|
| 7 |
"""
|
| 8 |
import os
|
|
|
|
|
|
|
| 9 |
from datetime import datetime
|
| 10 |
from zoneinfo import ZoneInfo
|
| 11 |
|
|
@@ -160,17 +162,30 @@ def _frontmatter(text):
|
|
| 160 |
|
| 161 |
# ---------- API-Endpunkte ----------
|
| 162 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 163 |
def analysieren_api(key, audio, num_speakers):
|
| 164 |
if not _pruefe(key):
|
| 165 |
return {"ok": False, "fehler": "Ungültiger Zugangsschlüssel."}
|
| 166 |
if not audio:
|
| 167 |
return {"ok": False, "fehler": "Kein Audio übermittelt."}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 168 |
try:
|
| 169 |
daten = _analyse_gpu(audio, num_speakers)
|
| 170 |
except Exception as e:
|
| 171 |
return {"ok": False, "fehler": f"Analyse fehlgeschlagen: {e}"}
|
| 172 |
zeitpunkt = datetime.now(ZEITZONE)
|
| 173 |
-
md_text = _markdown(daten,
|
| 174 |
try:
|
| 175 |
pfad = _speichern(md_text, zeitpunkt)
|
| 176 |
except Exception:
|
|
|
|
| 6 |
verbraucht.
|
| 7 |
"""
|
| 8 |
import os
|
| 9 |
+
import subprocess
|
| 10 |
+
import tempfile
|
| 11 |
from datetime import datetime
|
| 12 |
from zoneinfo import ZoneInfo
|
| 13 |
|
|
|
|
| 162 |
|
| 163 |
# ---------- API-Endpunkte ----------
|
| 164 |
|
| 165 |
+
def _nach_wav(pfad):
|
| 166 |
+
"""Konvertiert beliebige Audioformate (m4a, webm, …) nach 16-kHz-Mono-WAV."""
|
| 167 |
+
ziel = tempfile.NamedTemporaryFile(suffix=".wav", delete=False).name
|
| 168 |
+
subprocess.run(["ffmpeg", "-y", "-loglevel", "error", "-i", pfad,
|
| 169 |
+
"-ar", "16000", "-ac", "1", ziel], check=True)
|
| 170 |
+
return ziel
|
| 171 |
+
|
| 172 |
+
|
| 173 |
def analysieren_api(key, audio, num_speakers):
|
| 174 |
if not _pruefe(key):
|
| 175 |
return {"ok": False, "fehler": "Ungültiger Zugangsschlüssel."}
|
| 176 |
if not audio:
|
| 177 |
return {"ok": False, "fehler": "Kein Audio übermittelt."}
|
| 178 |
+
quelle = os.path.basename(audio)
|
| 179 |
+
try:
|
| 180 |
+
audio = _nach_wav(audio)
|
| 181 |
+
except subprocess.CalledProcessError:
|
| 182 |
+
return {"ok": False, "fehler": "Audioformat konnte nicht gelesen werden."}
|
| 183 |
try:
|
| 184 |
daten = _analyse_gpu(audio, num_speakers)
|
| 185 |
except Exception as e:
|
| 186 |
return {"ok": False, "fehler": f"Analyse fehlgeschlagen: {e}"}
|
| 187 |
zeitpunkt = datetime.now(ZEITZONE)
|
| 188 |
+
md_text = _markdown(daten, quelle, zeitpunkt)
|
| 189 |
try:
|
| 190 |
pfad = _speichern(md_text, zeitpunkt)
|
| 191 |
except Exception:
|