WIP: parakeeet

stt_voice_messages/helper.py

@@ -13,16 +13,53 @@
 # You should have received a copy of the GNU General Public License
 # along with Gajim. If not, see <http://www.gnu.org/licenses/>.
 
+import logging
+import typing
 from dataclasses import dataclass
+from pathlib import Path
 
+import gi
+import numpy as np
 from gi.repository import Gio, GObject
 
+try:
+    gi.require_version('Gst', '1.0')
+    from gi.repository import Gst
+except Exception:
+    if typing.TYPE_CHECKING:
+        from gi.repository import Gst
+
+log = logging.getLogger('gajim.p.sttvm_helper')
+
 
 @dataclass
 class Results:
     text: str
 
 
+def load_audio(path: Path, sample_rate: int = 16000) -> np.ndarray:
+    Gst.init(None)
+    pipeline = Gst.parse_launch(
+        'filesrc name=src ! decodebin ! audioconvert ! audioresample ! '
+        f'audio/x-raw,format=F32LE,rate={sample_rate},channels=1 ! '
+        'appsink name=sink sync=false'
+    )
+    pipeline.get_by_name('src').set_property('location', str(path))
+    sink = pipeline.get_by_name('sink')
+    chunks: list[np.ndarray] = []
+
+    pipeline.set_state(Gst.State.PLAYING)
+    while (sample := sink.emit('try-pull-sample', 10 * Gst.SECOND)) is not None:
+        buf = sample.get_buffer()
+        _, info = buf.map(Gst.MapFlags.READ)
+        chunks.append(np.frombuffer(bytes(info.data), dtype=np.float32))
+        buf.unmap(info)
+    pipeline.set_state(Gst.State.NULL)
+
+    if not chunks:
+        raise RuntimeError(f'Could not decode audio: {path}')
+    return np.concatenate(chunks)
+
 
 '''
 https://discourse.gnome.org/t/gtk-threading-problem-with-glib-idle-add/13597/5
@@ -57,6 +94,7 @@ class BackgroundTask(GObject.Object):
             retval = self.function()
             task.return_value(retval)
         except Exception as e:
+            log.exception('Background task failed')
             task.return_value(e)
 
     def finish(self):