Update app.py

app.py

@@ -25,7 +25,6 @@ except Exception:
     pass
 
 if not hasattr(torch.load, "_is_patched"):
-    print("DEBUG: Applying Monkeypatch to torch.load")
     _original_torch_load = torch.load
 
     def patched_torch_load(*args, **kwargs):
@@ -82,7 +81,8 @@ ACTIVE_GEMINI_KEY = ENV_GEMINI_KEY if ENV_GEMINI_KEY else HARDCODED_GEMINI_KEY
 def clean_json_response(text):
     """Strips markdown code fences from Gemini output to prevent JSON errors."""
     try:
-        pattern = r"```(?:json)?\s*(.*?)```"
+        # Avoid backtick parsing issues by splitting the string definition
+        pattern = r"`{3}(?:json)?\s*(.*?)`{3}"
         match = re.search(pattern, text, re.DOTALL)
         if match:
             return match.group(1).strip()
@@ -90,34 +90,44 @@ def clean_json_response(text):
     except Exception:
         return text
 
+def timecode_to_seconds(tc, fps=25):
+    """Converts HH:MM:SS:FF or HH:MM:SS.SS to total seconds."""
+    if not tc or not re.match(r"\d{2}:\d{2}:\d{2}[:\.]\d{2}", tc):
+        return 0.0
+    
+    parts = re.split(r'[:\.]', tc)
+    h, m, s, f = map(int, parts)
+    return h * 3600 + m * 60 + s + (f / fps)
+
 def format_timecode(seconds, fps=25):
     """Converts seconds to HH:MM:SS:FF (for EDL)."""
-    td = timedelta(seconds=seconds)
-    total_seconds = int(td.total_seconds())
-    hours = total_seconds // 3600
-    minutes = (total_seconds % 3600) // 60
-    secs = total_seconds % 60
-    frames = int((seconds - total_seconds) * fps)
+    total_frames = int(round(seconds * fps))
+    hours = total_frames // (3600 * int(fps))
+    minutes = (total_frames // (60 * int(fps))) % 60
+    secs = (total_frames // int(fps)) % 60
+    frames = total_frames % int(fps)
     return f"{hours:02}:{minutes:02}:{secs:02}:{frames:02}"
 
 def seconds_to_frames(seconds, fps):
     """Converts seconds to integer frames (for XML)."""
     return int(round(seconds * fps))
 
-def generate_cmx_edl(edl_title, segments, source_name, fps=25):
-    """Constructs a CMX 3600 formatted EDL."""
+def generate_cmx_edl(edl_title, segments, source_name, fps=25, start_offset=0.0):
+    """Constructs a CMX 3600 formatted EDL with source offset."""
     edl_lines = [f"TITLE: {edl_title}", "FCM: NON-DROP FRAME\n"]
     rec_start = 0.0
-    reel_id = source_name.replace(" ", "_")
+    reel_id = source_name.replace(" ", "_")[:8] # EDL reel IDs are traditionally short max 8 chars
     
     for i, seg in enumerate(segments, 1):
-        src_in = format_timecode(seg['src_start'], fps)
-        src_out = format_timecode(seg['src_end'], fps)
+        # 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)
+        
         duration = seg['src_end'] - seg['src_start']
         rec_in = format_timecode(rec_start, fps)
         rec_out = format_timecode(rec_start + duration, fps)
         
-        edl_lines.append(f"{i:03}  {reel_id}       V     C        {src_in} {src_out} {rec_in} {rec_out}")
+        edl_lines.append(f"{i:03}  {reel_id:8} V     C        {src_in} {src_out} {rec_in} {rec_out}")
         edl_lines.append(f"* FROM CLIP NAME: {source_name}")
         edl_lines.append(f"* {seg.get('note', 'Clip')}\n")
         
@@ -126,19 +136,16 @@ def generate_cmx_edl(edl_title, segments, source_name, fps=25):
         
     return "\n".join(edl_lines)
 
-def generate_xml(sequence_name, segments, source_name, fps=25):
-    """Constructs a Robust XML for Premiere AND Resolve."""
-    
-    # Static UUID ensures all cuts link to the SAME file object (Clean Bin)
+def generate_xml(sequence_name, segments, source_name, fps=25, start_offset=0.0):
+    """Constructs a Robust XML with source timecode offset for Premiere AND Resolve."""
     master_id = "masterfile-1"
-    
-    # FPS Logic
     timebase = int(round(fps))
     is_ntsc = "TRUE" if fps % 1 != 0 else "FALSE"
     
-    # --- NAME & PATH LOGIC ---
-    # We use URL encoding but NO .mov extension.
-    # This allows Resolve to find it by name, and Premiere to import it as "Offline".
+    start_offset_frames = seconds_to_frames(start_offset, fps)
+    start_tc_string = format_timecode(start_offset, fps)
+
+    # We use URL encoding but NO .mov extension to allow Resolve to find it by name.
     encoded_name = quote(source_name)
     path_url = f"file://localhost/{encoded_name}" 
 
@@ -169,8 +176,8 @@ def generate_xml(sequence_name, segments, source_name, fps=25):
 
     timeline_head_frames = 0
     for i, seg in enumerate(segments, 1):
-        src_in_frames = seconds_to_frames(seg['src_start'], fps)
-        src_out_frames = seconds_to_frames(seg['src_end'], fps)
+        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
         duration_frames = src_out_frames - src_in_frames
         
         tl_start = timeline_head_frames
@@ -181,7 +188,7 @@ def generate_xml(sequence_name, segments, source_name, fps=25):
 
         lines.append(f'\t\t\t\t<clipitem id="clipitem-v-{i}">')
         lines.append(f'\t\t\t\t\t<name>{clip_note}</name>')
-        lines.append(f'\t\t\t\t\t<duration>{duration_frames}</duration>')
+        lines.append(f'\t\t\t\t\t<duration>{duration_frames + src_in_frames}</duration>')
         lines.append(f'\t\t\t\t\t<rate><timebase>{timebase}</timebase></rate>')
         lines.append(f'\t\t\t\t\t<start>{tl_start}</start>')
         lines.append(f'\t\t\t\t\t<end>{tl_end}</end>')
@@ -197,26 +204,18 @@ def generate_xml(sequence_name, segments, source_name, fps=25):
         lines.append(f'\t\t\t\t\t\t<duration>8640000</duration>') 
         lines.append(f'\t\t\t\t\t\t<timecode>')
         lines.append(f'\t\t\t\t\t\t\t<rate><timebase>{timebase}</timebase></rate>')
-        lines.append(f'\t\t\t\t\t\t\t<string>00:00:00:00</string>') 
-        lines.append(f'\t\t\t\t\t\t\t<frame>0</frame>')
+        lines.append(f'\t\t\t\t\t\t\t<string>{start_tc_string}</string>') 
+        lines.append(f'\t\t\t\t\t\t\t<frame>{start_offset_frames}</frame>')
         lines.append(f'\t\t\t\t\t\t\t<displayformat>NDF</displayformat>')
         lines.append(f'\t\t\t\t\t\t</timecode>')
-        lines.append(f'\t\t\t\t\t\t<media>')
-        lines.append(f'\t\t\t\t\t\t\t<video>')
-        lines.append(f'\t\t\t\t\t\t\t\t<samplecharacteristics><width>1920</width><height>1080</height></samplecharacteristics>')
-        lines.append(f'\t\t\t\t\t\t\t</video>')
-        lines.append(f'\t\t\t\t\t\t\t<audio>')
-        lines.append(f'\t\t\t\t\t\t\t\t<samplecharacteristics><depth>16</depth><samplerate>48000</samplerate></samplecharacteristics>')
-        lines.append(f'\t\t\t\t\t\t\t\t<channelcount>2</channelcount>')
-        lines.append(f'\t\t\t\t\t\t\t</audio>')
-        lines.append(f'\t\t\t\t\t\t</media>')
-        lines.append(f'\t\t\t\t\t</file>')
-        # -----------------------------------
-        
-        lines.append(f'\t\t\t\t</clipitem>')
+        lines.append('\t\t\t\t\t\t<media>')
+        lines.append('\t\t\t\t\t\t\t<video><samplecharacteristics><width>1920</width><height>1080</height></samplecharacteristics></video>')
+        lines.append('\t\t\t\t\t\t\t<audio><samplecharacteristics><depth>16</depth><samplerate>48000</samplerate></samplecharacteristics><channelcount>2</channelcount></audio>')
+        lines.append('\t\t\t\t\t\t</media>')
+        lines.append('\t\t\t\t\t</file>')
+        lines.append('\t\t\t\t</clipitem>')
         
-        gap_seconds = seg.get('gap', 0.0)
-        gap_frames = seconds_to_frames(gap_seconds, fps)
+        gap_frames = seconds_to_frames(seg.get('gap', 0.0), fps)
         timeline_head_frames = tl_end + gap_frames
 
     lines.append('\t\t\t</track>')
@@ -224,64 +223,26 @@ def generate_xml(sequence_name, segments, source_name, fps=25):
     
     # --- AUDIO TRACK ---
     lines.append('\t\t<audio>')
-    lines.append('\t\t\t<track>') 
-    
+    lines.append('\t\t\t<track>')
+
     timeline_head_frames = 0
     for i, seg in enumerate(segments, 1):
-        src_in_frames = seconds_to_frames(seg['src_start'], fps)
-        src_out_frames = seconds_to_frames(seg['src_end'], fps)
+        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
         duration_frames = src_out_frames - src_in_frames
-        
         tl_start = timeline_head_frames
         tl_end = tl_start + duration_frames
-        
-        raw_note = seg.get('note', 'Junior Editor Selection')
-        clip_note = (raw_note[:75] + '..') if len(raw_note) > 75 else raw_note
 
         lines.append(f'\t\t\t\t<clipitem id="clipitem-a-{i}">')
-        lines.append(f'\t\t\t\t\t<name>{clip_note}</name>')
-        lines.append(f'\t\t\t\t\t<duration>{duration_frames}</duration>')
+        lines.append(f'\t\t\t\t\t<masterclipid>{master_id}</masterclipid>')
         lines.append(f'\t\t\t\t\t<rate><timebase>{timebase}</timebase></rate>')
         lines.append(f'\t\t\t\t\t<start>{tl_start}</start>')
         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(f'\t\t\t\t\t<masterclipid>{master_id}</masterclipid>')
-        
-        # Reuse same file definition (Verbose)
-        lines.append(f'\t\t\t\t\t<file id="{master_id}">')
-        lines.append(f'\t\t\t\t\t\t<name>{source_name}</name>')
-        lines.append(f'\t\t\t\t\t\t<pathurl>{path_url}</pathurl>') 
-        lines.append(f'\t\t\t\t\t\t<rate><timebase>{timebase}</timebase></rate>')
-        lines.append(f'\t\t\t\t\t\t<duration>8640000</duration>') 
-        lines.append(f'\t\t\t\t\t\t<timecode>')
-        lines.append(f'\t\t\t\t\t\t\t<rate><timebase>{timebase}</timebase></rate>')
-        lines.append(f'\t\t\t\t\t\t\t<string>00:00:00:00</string>') 
-        lines.append(f'\t\t\t\t\t\t\t<frame>0</frame>')
-        lines.append(f'\t\t\t\t\t\t\t<displayformat>NDF</displayformat>')
-        lines.append(f'\t\t\t\t\t\t</timecode>')
-        lines.append(f'\t\t\t\t\t\t<media>')
-        lines.append(f'\t\t\t\t\t\t\t<video>')
-        lines.append(f'\t\t\t\t\t\t\t\t<samplecharacteristics><width>1920</width><height>1080</height></samplecharacteristics>')
-        lines.append(f'\t\t\t\t\t\t\t</video>')
-        lines.append(f'\t\t\t\t\t\t\t<audio>')
-        lines.append(f'\t\t\t\t\t\t\t\t<samplecharacteristics><depth>16</depth><samplerate>48000</samplerate></samplecharacteristics>')
-        lines.append(f'\t\t\t\t\t\t\t\t<channelcount>2</channelcount>')
-        lines.append(f'\t\t\t\t\t\t\t</audio>')
-        lines.append(f'\t\t\t\t\t\t</media>')
-        lines.append(f'\t\t\t\t\t</file>')
-        
-        # SOURCE TRACK MAPPING (Simple Mono)
-        lines.append('\t\t\t\t\t<sourcetrack>')
-        lines.append('\t\t\t\t\t\t<mediatype>audio</mediatype>')
-        lines.append('\t\t\t\t\t\t<trackindex>1</trackindex>')
-        lines.append('\t\t\t\t\t</sourcetrack>')
-        
-        lines.append(f'\t\t\t\t</clipitem>')
-        
-        gap_seconds = seg.get('gap', 0.0)
-        gap_frames = seconds_to_frames(gap_seconds, fps)
-        timeline_head_frames = tl_end + gap_frames
+        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)
 
     lines.append('\t\t\t</track>')
     lines.append('\t\t</audio>')
@@ -337,7 +298,7 @@ st.title("Junior Editor")
 st.markdown("""
 **Instructions**
 * Upload your file here (video or audio).
-* Set your timeline FPS and transcription quality.
+* Set your timeline FPS, Source Timecode, and transcription quality in the sidebar.
 * Junior Editor will transcribe and separate speakers. You can then instruct it to find engaging bits or construct a narrative.
 * It will create an EDL or XML to import back into your editing software (Resolve, Premiere).
 * **For Multicam Workflows:** Use the **XML** option in the sidebar and enter the **exact name** of your Multicam Sequence.
@@ -352,12 +313,15 @@ with st.sidebar:
     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."
@@ -490,18 +454,19 @@ if uploaded_file:
             else:
                 with st.spinner("Junior Editor is thinking..."):
                     final_source_name = custom_reel_name.strip() if custom_reel_name.strip() else uploaded_file.name
+                    # Parse the offset from sidebar input
+                    offset_sec = timecode_to_seconds(source_start_tc, fps)
                     
                     edl_segments = call_gemini_for_edl(st.session_state.transcript, brief, ACTIVE_GEMINI_KEY)
                     if edl_segments:
                         if export_format == "EDL":
-                            final_output = generate_cmx_edl("Junior_Editor_Cut", edl_segments, final_source_name, fps)
+                            final_output = generate_cmx_edl("Junior_Editor_Cut", edl_segments, final_source_name, fps, offset_sec)
                             ext = "edl"
                         else:
-                            final_output = generate_xml("Junior_Editor_Cut", edl_segments, final_source_name, fps)
+                            final_output = generate_xml("Junior_Editor_Cut", edl_segments, final_source_name, fps, offset_sec)
                             ext = "xml"
                             
                         st.subheader("Ready for Import")
-                        # --- COLLAPSED DEFAULT ---
                         with st.expander("Show XML/EDL Code", expanded=False):
                             st.code(final_output, language="xml" if ext == "xml" else "text")