add external metadata input and group inputs

app.py

@@ -26,11 +26,6 @@ def safe_read(i: int, a: list):
     else:
         return a[i]
 
-def getTrigger(ad: int, a: list, max: int = 1024) -> int:
-    i = ad
-    while not (a[i] < 126 and not a[i + 4] < 130 or i - ad > max):
-        i += 1
-    return i
 
 def getRenderCords(ta: list, idx: int, res: int = 1024, size: tuple = (1280, 720)) -> list:
     i = idx - res // 2
@@ -54,14 +49,13 @@ def totopleft(coord, width=1280, height=720):
 
 def getTrigger(ad: int, a: list, max: int = 1024) -> int:
     i = ad
-    while not (safe_read(i,a) < 124 and safe_read(i+2,a) < 128 or i - ad > max):
+    while not (safe_read(i,a) < 126 and safe_read(i+6,a) < 130 or i - ad > max):
         i += 1
     return i
 
 def extract_cover_image(mp3_file):
     audio = MP3(mp3_file, ID3=ID3)
     if audio.tags == None:
-        
         return -1
     for tag in audio.tags.values():
         if isinstance(tag, APIC):
@@ -73,8 +67,8 @@ def extract_cover_image(mp3_file):
 
 def getTitleAndArtist(mp3_file):
     audio = MP3(mp3_file, ID3=ID3)
-    title = audio.get('TIT2', TIT2(encoding=3, text='Unknown Title')).text[0]
-    artist = audio.get('TPE1', TPE1(encoding=3, text='Unknown Artist')).text[0]
+    title = audio.get('TIT2', TIT2(encoding=3, text='')).text[0]
+    artist = audio.get('TPE1', TPE1(encoding=3, text='')).text[0]
         
     
     return title, artist
@@ -176,7 +170,7 @@ def stripinvisibles(s):
         e.replace(i,"")
     return e
 
-def main(file, name, fps=30, res: tuple=(1280,720), oscres=512, sr=11025, lyrics=None):
+def main(file, name, fps=30, res: tuple=(1280,720), oscres=512, sr=11025, lyrics=None, img=None, tit=None, ast=None):
     p = gr.Progress()
     LRC2SRT.clear()
     if os.path.exists("out.srt"):
@@ -201,7 +195,7 @@ def main(file, name, fps=30, res: tuple=(1280,720), oscres=512, sr=11025, lyrics
                 gr.Warning("Lyrics file is invalid, skipping")
         except Exception as e:
             print(traceback.format_exc())
-            gr.Warning("Failed to parse lyrics, ensure there are no blank lines in between and invisible characters")
+            gr.Warning("Failed to parse lyrics, ensure there are no blank lines in between")
         
     os.makedirs(path+f'out/{name}/', exist_ok=True)
     global iii
@@ -214,18 +208,25 @@ def main(file, name, fps=30, res: tuple=(1280,720), oscres=512, sr=11025, lyrics
     samples_array = y_u8.tolist()
     p(0.5,"extracting metadata")
     # Extract cover image, title, and artist
+    cover_file = None
+    if img:
+        cover_file = Image.open(img)
     cover_img = extract_cover_image(audio_path)
-    if cover_img is None:
-        raise gr.Error("Mp3 must have a cover image")
-        return  # Exit if no cover image found
-    elif cover_img == -1:
-        raise gr.Error("Mp3 is missing tags")
-        return
+    if cover_img == None:
+        if img:
+            cover_img = cover_file
+        else:
+            raise gr.Error("Mp3 must have a cover image, upload the image under the 'Metadata' section")
+    elif cover_img == -1 and not (tit or ast):
+        raise gr.Error("Mp3 is missing tags, add the info under the 'Metadata' section")
         
 
     title, artist = getTitleAndArtist(audio_path) 
-    if title == 'Unknown Title' or artist == 'Unknown Artist':
-        gr.Warning('Missing Title or Artist')
+    if title == '' or artist == '':
+        if not (tit or ast):
+            gr.Warning('Missing Title or Artist')
+        else:
+            title, artist = tit, ast
     dominant_color = getColour(cover_img)
 
     # Frame rendering parameters
@@ -247,7 +248,7 @@ def main(file, name, fps=30, res: tuple=(1280,720), oscres=512, sr=11025, lyrics
 
     except Exception as e:
         print('Ended in error: ' + traceback.format_exc(), iii)
-        #gr.Info("Rendering had errored, this typically an out of range error")
+        gr.e
     p = gr.Progress()
     p(0.5,desc="Compiling video")
     print('FFMPEG')
@@ -281,28 +282,53 @@ def main(file, name, fps=30, res: tuple=(1280,720), oscres=512, sr=11025, lyrics
     ]
     subprocess.run(ffmpeg_cmd)
 
-def gradio_interface(audio_file, lyrics, output_name, fps=30, vidwidth=1280, vidheight=720, oscres=512, sr=11025):
+def gradio_interface(audio_file, lyrics, output_name, fps=30, vidwidth=1280, vidheight=720, oscres=512, img=None, tit=None, ast=None):
     resolution = f"{vidwidth}x{vidheight}"
     res = tuple(map(int, resolution.split('x')))
-    main(audio_file, output_name, fps=fps, res=res, oscres=oscres, sr=sr, lyrics=lyrics)
+    main(audio_file, output_name, fps=fps, res=res, oscres=oscres, lyrics=lyrics, img=img, tit=tit, ast=ast)
     time.sleep(5)
     
     shutil.rmtree("out")
     return f"{output_name}.mp4"
 
-# Define Gradio interface with progress bar
+# Define Gradio interface with accordions
+inputs = [
+    gr.Accordion(
+        title="Audio Settings",
+        items=[
+            gr.components.File(label="Upload your MP3 file", file_count='single', file_types=['mp3']),
+            gr.components.File(label="(Optional) Upload Lyrics as LRC or SRT", file_count='single', file_types=['lrc','srt'])
+        ],
+        open=True
+    ),
+    gr.Accordion(
+        title="Video Output Settings",
+        items=[
+            gr.components.Textbox(label="Output Video Name", value='video'),
+            gr.components.Slider(label="Frames per Second", minimum=20, maximum=60, step=1, value=30),
+            gr.components.Slider(label="Output Video Width", minimum=100, maximum=2000, value=1280, step=2),
+            gr.components.Slider(label="Output Video Height", minimum=100, maximum=2000, value=720, step=2)
+        ]
+    ),
+    gr.Accordion(
+        title="Advanced Options",
+        items=[
+            gr.components.Slider(label="Number of Visualization Segments", minimum=256, maximum=2048, step=2, value=512),
+        ]
+    ),
+    gr.Accordion(
+        title="Mp3 Metadata (Use if mp3 does not have tags)",
+        items=[
+            gr.components.Image(label='Cover Art'),
+            gr.components.Textbox(label='Title'),
+            gr.components.Textbox(label='Artists')
+        ]
+    )
+]
+
 iface = gr.Interface(
     fn=gradio_interface,
-    inputs=[
-        gr.components.File(label="Upload your MP3 file", file_count='single', file_types=['mp3']),
-        gr.components.File(label="(Optional) Upload Lyrics as LRC or SRT", file_count='single', file_types=['lrc','srt']),
-        gr.components.Textbox(label="Output Video Name", value='video'),
-        gr.components.Slider(label="Frames per Second", minimum=20, maximum=60, step=1, value=30),
-        gr.components.Slider(label="Output Video Width", minimum=100, maximum=2000, value=1280, step=2),
-        gr.components.Slider(label="Output Video Height", minimum=100, maximum=2000, value=720, step=2),
-        gr.components.Slider(label="Number of Visualization Segments", minimum=256, maximum=2048, step=2, value=512),
-        #gr.components.Slider(label="Scope Sample Rate", minimum=8000, maximum=44100, step=5, value=11025)
-    ],
+    inputs=inputs,
     outputs=gr.components.Video(label="Output"),
     title="MP3 to Video Visualization",
     description=""" Upload an MP3 file and configure parameters to create a visualization video.
@@ -311,4 +337,4 @@ iface = gr.Interface(
 )
 
 # Launch Gradio interface
-iface.launch()
+iface.launch()