fix: st.audio in separate.py and prepare_samples script

app/pages/Separate.py

@@ -81,7 +81,7 @@ def show_results(model_name: str, dir_name_output: str, file_sources: List):
                 )
             with cols[1]:
                 st_local_audio(pathname, key=f"output_{file}_{dir_name_output}")
-    log.info(f"Displaying results for {dir_name_output}")
+    log.info(f"Displaying results for {dir_name_output} - {model_name}")
 
 
 def body():
@@ -140,10 +140,7 @@ def body():
                 help="Supported formats: mp3, wav, ogg, flac.",
             )
             if uploaded_file is not None:
-                with open(in_path / uploaded_file.name, "wb") as f:
-                    f.write(uploaded_file.getbuffer())
-                filename = uploaded_file.name
-                st_local_audio(in_path / filename, key="input_upload_file")
+                st.audio(uploaded_file)
 
         elif option == "From URL":
             url = st.text_input(

scripts/inference.py

@@ -16,20 +16,20 @@ def main():
     p.add_argument("--pretrained_model", "-P", type=str, default="baseline.pth")
     p.add_argument("--input", "-i", required=True)
     p.add_argument("--output_dir", "-o", type=str, default="")
-    p.add_argument("--only_no_vocals", "-n", default=True)
+    p.add_argument("--full_mode", "-n", default=False)
     args = p.parse_args()
-
+    print(args)
     input_file = args.input
-
+    full_mode = bool(args.full_mode)
     model, device = load_model(pretrained_model=args.pretrained_model)
-    separate(
-        input=input_file,
-        model=model,
-        device=device,
-        output_dir=args.output_dir,
-        only_no_vocals=args.only_no_vocals,
-    )
-    if not args.only_no_vocals:
+
+    if full_mode:
+        separate(
+            input=input_file,
+            model=model,
+            device=device,
+            output_dir=args.output_dir,
+        )
         for stem, model_name in [("vocals", "htdemucs"), (None, "htdemucs"), (None, "htdemucs_6s")]:
             separator(
                 tracks=[Path(input_file)],
@@ -45,6 +45,14 @@ def main():
                 mp3_bitrate=320,
                 verbose=False,
             )
+    else:
+        separate(
+            input=input_file,
+            model=model,
+            device=device,
+            output_dir=args.output_dir,
+            only_no_vocals=True,
+        )
 
 
 if __name__ == "__main__":

scripts/prepare_samples.sh

@@ -1,5 +1,5 @@
 #!/bin/bash
-
+echo "Starting prepare_samples.sh..."
 # Read the secret into a variable
 export PREPARE_SAMPLES=$(cat /run/secrets/PREPARE_SAMPLES)
 
@@ -33,13 +33,15 @@ json_separate=$(cat separate_songs.json)
 
 # Iterate through keys and values
 for name in $(echo "${json_separate}" | jq -r 'keys[]'); do
-    url=$(echo "${json_separate}" | jq -r --arg name "${name}" '.[$name]')
-    echo "Separating ${name} from ${url}"
+    url=$(echo "${json_separate}" | jq -r --arg name "${name}" '.[$name][0]')
+    start_time=$(echo "${json_separate}" | jq -r --arg name "${name}" '.[$name][1]')
+    end_time=$(expr $start_time + 20)
+    echo "Separating ${name} from ${url} with start_time ${start_time} sec"
 
     # Download with pytube
-    yt-dlp ${url} -o "/tmp/${name}" --format "bestaudio/best"  --download-sections "*45-110"
+    yt-dlp ${url} -o "/tmp/${name}" --format "bestaudio/best"  --download-sections "*${start_time}-${end_time}"
 
     # Run inference
-    python inference.py --input /tmp/${name} --output /tmp --only_no_vocals false
+    python inference.py --input /tmp/${name} --output /tmp --full_mode 1
     echo "Done separating ${name}"
 done

scripts/sample_songs.json

@@ -1,8 +1,3 @@
 {
-    "dancing_queen": "https://www.youtube.com/watch?v=3qiMJt-JBb4",
-    "bohemian_rhapsody": "https://www.youtube.com/watch?v=yk3prd8GER4",
-    "i_want_it_that_way": "https://www.youtube.com/watch?v=qjlVAsvQLM8",
-    "let_it_be": "https://www.youtube.com/watch?v=FIV73iG_e5I",
-    "viva_la_vida": "https://www.youtube.com/watch?v=a1EYnngNHIA",
-    "zombie": "https://www.youtube.com/watch?v=8sM-rm4lFZg"
+    "dancing_queen": "https://www.youtube.com/watch?v=3qiMJt-JBb4"
 }

scripts/separate_songs.json

@@ -1,8 +1,8 @@
 {
-    "ABBA_-_Dancing_Queen": "https://www.youtube.com/watch?v=3qiMJt-JBb4",
-    "Queen_–_Bohemian_Rhapsody": "https://www.youtube.com/watch?v=yk3prd8GER4",
-    "Backstreet_Boys_-_I_Want_It_That_Way": "https://www.youtube.com/watch?v=qjlVAsvQLM8",
-    "The_Beatles_-_Let_It_Be": "https://www.youtube.com/watch?v=FIV73iG_e5I",
-    "Coldplay_-_Viva_La_Vida": "https://www.youtube.com/watch?v=a1EYnngNHIA",
-    "The_Cranberries_-_Zombie": "https://www.youtube.com/watch?v=8sM-rm4lFZg"
+    "ABBA_-_Dancing_Queen": ["https://www.youtube.com/watch?v=3qiMJt-JBb4",85],
+    "Queen_–_Bohemian_Rhapsody": ["https://www.youtube.com/watch?v=yk3prd8GER4", 86],
+    "Backstreet_Boys_-_I_Want_It_That_Way": ["https://www.youtube.com/watch?v=qjlVAsvQLM8", 43],
+    "The_Beatles_-_Let_It_Be": ["https://www.youtube.com/watch?v=FIV73iG_e5I", 82],
+    "Coldplay_-_Viva_La_Vida": ["https://www.youtube.com/watch?v=a1EYnngNHIA", 7],
+    "The_Cranberries_-_Zombie": ["https://www.youtube.com/watch?v=8sM-rm4lFZg", 82],
 }