Update flask_api_full_song.py

flask_api_full_song.py

@@ -65,27 +65,40 @@ def infer(audio_path, tran, spk, wav_format, task_id):
         audio_data, audio_sr = slicer.chunks2audio(audio_path, chunks)
 
         audio = []
-        for (slice_tag, data) in audio_data:
-            print(f'#=====segment start, {round(len(data) / audio_sr, 3)}s======')
 
-            length = int(np.ceil(len(data) / audio_sr * svc_model.target_sample))
+        def process_chunk(chunk_data, audio_sr, slice_tag, svc_model, audio):
+            length = int(np.ceil(len(chunk_data) / audio_sr * svc_model.target_sample))
             if slice_tag:
                 print('jump empty segment')
                 _audio = np.zeros(length)
             else:
                 # padd
                 pad_len = int(audio_sr * 0.5)
-                data = np.concatenate([np.zeros([pad_len]), data, np.zeros([pad_len])])
+                chunk_data = np.concatenate([np.zeros([pad_len]), chunk_data, np.zeros([pad_len])])
                 raw_path = io.BytesIO()
-                soundfile.write(raw_path, data, audio_sr, format="wav")
+                soundfile.write(raw_path, chunk_data, audio_sr, format="wav")
                 raw_path.seek(0)
                 out_audio, out_audio_shape, out_sr = svc_model.infer(spk, tran, raw_path)
                 svc_model.clear_empty()
                 _audio = out_audio.cpu().numpy()
                 pad_len = int(svc_model.target_sample * 0.5)
                 _audio = _audio[pad_len:-pad_len]
-
             audio.extend(list(infer_tool.pad_array(_audio, length)))
+
+        for (slice_tag, data) in audio_data:
+            print(f'#=====segment start, {round(len(data) / audio_sr, 3)}s======')
+
+            # Check if the segment is longer than 30 seconds
+            segment_length = len(data) / audio_sr
+            if segment_length > 30:
+                # Split the segment into chunks of 30 seconds or less
+                num_chunks = int(np.ceil(segment_length / 30))
+                chunk_length = int(len(data) / num_chunks)
+                for i in range(num_chunks):
+                    chunk_data = data[i * chunk_length:(i + 1) * chunk_length]
+                    process_chunk(chunk_data, audio_sr, slice_tag, svc_model, audio)
+            else:
+                process_chunk(data, audio_sr, slice_tag, svc_model, audio)
         out_wav_path = "/tmp/" + audio_name
         soundfile.write(out_wav_path, audio, svc_model.target_sample, format=wav_format)