Update app.py

app.py

@@ -270,15 +270,48 @@ interface_file = gr.Interface(
     article=article,
 )
 
-def combined_output(bilibili_url, start_time, end_time, speaker_choice, pitch_shift, f0_check, cluster_ratio, noise_scale, inference_method):
-    # 根据你之前提供的参数调用predict_song_from_yt
-    ai_with_instrumental, ai_vocal = predict_song_from_yt(bilibili_url, start_time, end_time, speaker_choice, pitch_shift, f0_check, cluster_ratio, noise_scale, inference_method)
-    
-    # 假设extract_vocal_demucs需要B站URL，起始时间和结束时间作为输入
-    # 注意：你可能需要调整这一部分，以确保你为extract_vocal_demucs提供了正确的参数
-    vocal_wav, instrumental_wav = extract_vocal_demucs(vocal_wav, instrumental_wav)
+def combined_output(
+    ytid_or_url,
+    start,
+    end,
+    speaker=speakers[0],
+    transpose: int = 0,
+    auto_predict_f0: bool = False,
+    cluster_infer_ratio: float = 0,
+    noise_scale: float = 0.4,
+    f0_method: str = "dio",
+    db_thresh: int = -40,
+    pad_seconds: float = 0.5,
+    chunk_seconds: float = 0.5,
+    absolute_thresh: bool = False,
+):
+    # 调用原来的函数
+    full_song_output, cloned_vox_output = predict_song_from_yt(
+        ytid_or_url,
+        start,
+        end,
+        speaker,
+        transpose,
+        auto_predict_f0,
+        cluster_infer_ratio,
+        noise_scale,
+        f0_method,
+        db_thresh,
+        pad_seconds,
+        chunk_seconds,
+        absolute_thresh,
+    )
     
-    return ai_with_instrumental, ai_vocal, vocal_wav, instrumental_wav
+    # 这里我们直接提取人声和伴奏，因为它们已经在predict_song_from_yt中被提取
+    original_track_filepath = download_youtube_clip(
+        ytid_or_url, start, end, "track.wav", force=True,
+        url_base="" if ytid_or_url.startswith("http") else "https://www.youtube.com/watch?v="
+    )
+    vox_wav, inst_wav = extract_vocal_demucs(demucs_model, original_track_filepath)
+
+    # 返回所有输出
+    return full_song_output, cloned_vox_output, (model.target_sample, vox_wav), (model.target_sample, inst_wav)
+