Update app.py

app.py

@@ -120,11 +120,15 @@ def generate_cmx_edl(edl_title, segments, source_name, fps=25, start_offset=0.0)
     reel_id = source_name.replace(" ", "_")[:8] # EDL reel IDs are traditionally short max 8 chars
     
     for i, seg in enumerate(segments, 1):
+        # Fallback to 'start'/'end' if 'src_start'/'src_end' are missing, and force float
+        src_start_val = float(seg.get('src_start', seg.get('start', 0.0)))
+        src_end_val = float(seg.get('src_end', seg.get('end', 0.0)))
+        
         # Add offset to source time
-        src_in = format_timecode(seg['src_start'] + start_offset, fps)
-        src_out = format_timecode(seg['src_end'] + start_offset, fps)
+        src_in = format_timecode(src_start_val + start_offset, fps)
+        src_out = format_timecode(src_end_val + start_offset, fps)
         
-        duration = seg['src_end'] - seg['src_start']
+        duration = src_end_val - src_start_val
         rec_in = format_timecode(rec_start, fps)
         rec_out = format_timecode(rec_start + duration, fps)
         
@@ -132,7 +136,7 @@ def generate_cmx_edl(edl_title, segments, source_name, fps=25, start_offset=0.0)
         edl_lines.append(f"* FROM CLIP NAME: {source_name}")
         edl_lines.append(f"* {seg.get('note', 'Clip')}\n")
         
-        gap = seg.get('gap', 0.0)
+        gap = float(seg.get('gap', 0.0))
         rec_start += duration + gap
         
     return "\n".join(edl_lines)
@@ -177,8 +181,11 @@ def generate_xml(sequence_name, segments, source_name, fps=25, start_offset=0.0)
 
     timeline_head_frames = 0
     for i, seg in enumerate(segments, 1):
-        src_in_frames = seconds_to_frames(seg['src_start'], fps) + start_offset_frames
-        src_out_frames = seconds_to_frames(seg['src_end'], fps) + start_offset_frames
+        src_start_val = float(seg.get('src_start', seg.get('start', 0.0)))
+        src_end_val = float(seg.get('src_end', seg.get('end', 0.0)))
+        
+        src_in_frames = seconds_to_frames(src_start_val, fps) + start_offset_frames
+        src_out_frames = seconds_to_frames(src_end_val, fps) + start_offset_frames
         duration_frames = src_out_frames - src_in_frames
         
         tl_start = timeline_head_frames
@@ -216,7 +223,7 @@ def generate_xml(sequence_name, segments, source_name, fps=25, start_offset=0.0)
         lines.append('\t\t\t\t\t</file>')
         lines.append('\t\t\t\t</clipitem>')
         
-        gap_frames = seconds_to_frames(seg.get('gap', 0.0), fps)
+        gap_frames = seconds_to_frames(float(seg.get('gap', 0.0)), fps)
         timeline_head_frames = tl_end + gap_frames
 
     lines.append('\t\t\t</track>')
@@ -228,8 +235,11 @@ def generate_xml(sequence_name, segments, source_name, fps=25, start_offset=0.0)
 
     timeline_head_frames = 0
     for i, seg in enumerate(segments, 1):
-        src_in_frames = seconds_to_frames(seg['src_start'], fps) + start_offset_frames
-        src_out_frames = seconds_to_frames(seg['src_end'], fps) + start_offset_frames
+        src_start_val = float(seg.get('src_start', seg.get('start', 0.0)))
+        src_end_val = float(seg.get('src_end', seg.get('end', 0.0)))
+        
+        src_in_frames = seconds_to_frames(src_start_val, fps) + start_offset_frames
+        src_out_frames = seconds_to_frames(src_end_val, fps) + start_offset_frames
         duration_frames = src_out_frames - src_in_frames
         tl_start = timeline_head_frames
         tl_end = tl_start + duration_frames
@@ -241,9 +251,9 @@ def generate_xml(sequence_name, segments, source_name, fps=25, start_offset=0.0)
         lines.append(f'\t\t\t\t\t<end>{tl_end}</end>')
         lines.append(f'\t\t\t\t\t<in>{src_in_frames}</in>')
         lines.append(f'\t\t\t\t\t<out>{src_out_frames}</out>')
-        lines.append('\t\t\t\t\t<sourcetrack><mediatype>audio</mediatype><trackindex>1</trackindex></souretrack>')
+        lines.append('\t\t\t\t\t<sourcetrack><mediatype>audio</mediatype><trackindex>1</trackindex></sourcetrack>')
         lines.append('\t\t\t\t</clipitem>')
-        timeline_head_frames = tl_end + seconds_to_frames(seg.get('gap', 0.0), fps)
+        timeline_head_frames = tl_end + seconds_to_frames(float(seg.get('gap', 0.0)), fps)
 
     lines.append('\t\t\t</track>')
     lines.append('\t\t</audio>')
@@ -357,52 +367,47 @@ st.markdown("""
 st.divider()
 
 with st.sidebar:
-    st.header("Project Settings")
-    fps_options = [23.98, 24, 25, 29.97, 30, 50, 59.94, 60]
-    fps = st.selectbox("Timeline FPS", fps_options, index=2)
-    
-    st.header("Source Timecode")
-    st.info("Set this if your source file should reference a specific time-of-day timecode (e.g. for MP3s).")
-    source_start_tc = st.text_input("Source Start Timecode", value="00:00:00:00", help="Format: HH:MM:SS:FF")
-    
-    st.header("Export Settings")
-    export_format = st.radio("Output Format", ["EDL", "XML (Multicam)"], index=0)
-    
-    input_label = "EDL Reel Name"
-    input_help = "Leave empty to use the uploaded file name."
-    if export_format == "XML (Multicam)":
-        input_label = "Multicam Sequence Name"
-        input_help = "EXACT name of your Multicam Clip in Resolve."
-    
-    st.info("💡 **Conform Helper**")
-    custom_reel_name = st.text_input(
-        input_label, 
-        placeholder="e.g. Interview_Day1_Multi", 
-        help=input_help
-    )
+    # Use containers so we can render Clip Settings visually above Project Settings,
+    # but still evaluate Export Settings first (needed to determine dynamic text labels).
+    clip_settings_container = st.container()
+    project_settings_container = st.container()
+
+    with project_settings_container:
+        st.header("Project Settings")
+        export_format = st.radio("Output Format", ["EDL", "XML (Multicam)"], index=0)
+        
+        fps_options = [23.98, 24, 25, 29.97, 30, 50, 59.94, 60]
+        fps = st.selectbox("Timeline FPS", fps_options, index=2)
+        
+        language_map = {
+            "Auto-Detect": None,
+            "English": "en",
+            "Spanish": "es",
+            "French": "fr",
+            "German": "de",
+            "Italian": "it",
+            "Portuguese": "pt"
+        }
+        selected_lang_label = st.selectbox("Audio Language", list(language_map.keys()), index=1)
+        target_language = language_map[selected_lang_label]
+        
+        num_speakers = st.number_input("Speakers (0=Auto)", min_value=0, value=0)
 
-    st.header("Model Settings")
-    model_size = st.selectbox("Whisper Model", ["large-v2", "medium", "base"], index=0)
-    
-    language_map = {
-        "Auto-Detect": None,
-        "English": "en",
-        "Spanish": "es",
-        "French": "fr",
-        "German": "de",
-        "Italian": "it",
-        "Portuguese": "pt"
-    }
-    selected_lang_label = st.selectbox("Audio Language", list(language_map.keys()), index=1)
-    target_language = language_map[selected_lang_label]
-    
-    num_speakers = st.number_input("Speakers (0=Auto)", min_value=0, value=0)
-    
-    st.divider()
-    if ACTIVE_HF_TOKEN == "PASTE_YOUR_HF_TOKEN_HERE":
-        st.warning("⚠️ HF_TOKEN not set in Secrets!")
-    else:
-        st.success("✅ HF_TOKEN Loaded")
+    with clip_settings_container:
+        st.header("Clip Settings")
+        source_start_tc = st.text_input("Source Start Timecode", value="00:00:00:00", help="Format: HH:MM:SS:FF")
+        
+        input_label = "EDL Reel Name"
+        input_help = "Leave empty to use the uploaded file name."
+        if export_format == "XML (Multicam)":
+            input_label = "Multicam Sequence Name"
+            input_help = "EXACT name of your Multicam Clip in Resolve."
+        
+        custom_reel_name = st.text_input(
+            input_label, 
+            placeholder="e.g. Interview_Day1_Multi", 
+            help=input_help
+        )
 
 uploaded_file = st.file_uploader("Upload Video/Audio Clip", type=["mp4", "m4a", "wav", "mp3", "mov"])
 
@@ -437,9 +442,10 @@ if uploaded_file:
                     if device == "cpu": st.warning("⚠️ No GPU detected.")
                     
                     # Phase 2
-                    status_text.markdown(f"**Phase 2/4: Transcribing (Whisper {model_size})... This is the longest step.**")
+                    status_text.markdown("**Phase 2/4: Transcribing (Whisper)... This is the longest step.**")
                     compute_type = "float16" if device == "cuda" else "int8"
-                    model = whisperx.load_model(model_size, device, compute_type=compute_type)
+                    # Default model size hardcoded to 'large-v2' to keep high quality while removing the sidebar toggle
+                    model = whisperx.load_model("large-v2", device, compute_type=compute_type)
                     audio = whisperx.load_audio("temp_audio.wav")
                     result = model.transcribe(audio, batch_size=16, language=target_language)
                     del model