Update app.py

app.py

@@ -119,16 +119,19 @@ def generate_cmx_edl(edl_title, segments, source_name, fps=25, start_offset_fram
     reel_id = source_name.replace(" ", "_")[:8] # EDL reel IDs are traditionally short max 8 chars
     
     for i, seg in enumerate(segments, 1):
-        is_vo = seg.get('type') == 'vo' or 'text' in seg
+        seg_type = seg.get('type', 'clip')
+        is_vo = seg_type == 'vo' or ('text' in seg and seg_type != 'graphic')
+        is_graphic = seg_type == 'graphic'
         
-        if is_vo:
-            vo_text = seg.get('text', 'VO')
-            duration_sec = float(seg.get('duration', max(1.0, len(vo_text.split()) / 2.5)))
+        if is_vo or is_graphic:
+            text_content = seg.get('text', 'Placeholder')
+            default_dur = max(1.0, len(text_content.split()) / 2.5) if is_vo else 4.0
+            duration_sec = float(seg.get('duration', default_dur))
             duration_frames = seconds_to_frames(duration_sec, fps)
             src_in_frames = 0
             src_out_frames = duration_frames
-            current_source = "GEN_VO"
-            note_text = f"VO SCRIPT: {vo_text}"
+            current_source = "GEN_VO" if is_vo else "GEN_GFX"
+            note_text = f"{'VO SCRIPT' if is_vo else 'GRAPHIC'}: {text_content}"
         else:
             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)))
@@ -145,7 +148,7 @@ def generate_cmx_edl(edl_title, segments, source_name, fps=25, start_offset_fram
         rec_out = frames_to_timecode(rec_start_frames + duration_frames, fps)
         
         edl_lines.append(f"{i:03}  {current_source:8} V     C        {src_in} {src_out} {rec_in} {rec_out}")
-        if not is_vo:
+        if not is_vo and not is_graphic:
             edl_lines.append(f"* FROM CLIP NAME: {source_name}")
         edl_lines.append(f"* {note_text}\n")
         
@@ -192,26 +195,31 @@ def generate_xml(sequence_name, segments, source_name, fps=25, start_offset_fram
 
     file_defined = False
     vo_file_defined = False
+    gfx_file_defined = False
     timeline_head_frames = 0
     
     for i, seg in enumerate(segments, 1):
-        is_vo = seg.get('type') == 'vo' or 'text' in seg
+        seg_type = seg.get('type', 'clip')
+        is_vo = seg_type == 'vo' or ('text' in seg and seg_type != 'graphic')
+        is_graphic = seg_type == 'graphic'
         
-        if is_vo:
-            vo_text = seg.get('text', 'VO')
-            duration_sec = float(seg.get('duration', max(1.0, len(vo_text.split()) / 2.5)))
+        if is_vo or is_graphic:
+            text_content = seg.get('text', 'Placeholder')
+            default_dur = max(1.0, len(text_content.split()) / 2.5) if is_vo else 4.0
+            duration_sec = float(seg.get('duration', default_dur))
             duration_frames = seconds_to_frames(duration_sec, fps)
             src_in_frames = 0
             src_out_frames = duration_frames
-            master_id_to_use = "masterfile-vo"
-            source_name_to_use = "Generated_VO_Placeholder"
-            clip_note = f"VO SCRIPT: {vo_text}"
+            master_id_to_use = "masterfile-vo" if is_vo else "masterfile-gfx"
+            source_name_to_use = "Generated_VO_Placeholder" if is_vo else "Generated_Graphic_Placeholder"
+            prefix = "VO SCRIPT" if is_vo else "GRAPHIC CARD"
+            clip_note = f"{prefix}: {text_content}"
             clip_name_disp = (clip_note[:75] + '..') if len(clip_note) > 75 else clip_note
         else:
             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)
-            src_out_frames = seconds_to_frames(src_end_val, fps)
+            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
             master_id_to_use = master_id
             source_name_to_use = source_name
@@ -254,6 +262,27 @@ def generate_xml(sequence_name, segments, source_name, fps=25, start_offset_fram
                 vo_file_defined = True
             else:
                 lines.append(f'\t\t\t\t\t<file id="{master_id_to_use}"/>')
+        elif is_graphic:
+            if not gfx_file_defined:
+                lines.append(f'\t\t\t\t\t<file id="{master_id_to_use}">')
+                lines.append(f'\t\t\t\t\t\t<name>{source_name_to_use}</name>')
+                lines.append(f'\t\t\t\t\t\t<pathurl>file://localhost/{source_name_to_use}.mov</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('\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>')
+                # Intentionally omitting audio block for graphics placeholder to keep NLE track structure clean
+                lines.append('\t\t\t\t\t\t</media>')
+                lines.append('\t\t\t\t\t</file>')
+                gfx_file_defined = True
+            else:
+                lines.append(f'\t\t\t\t\t<file id="{master_id_to_use}"/>')
         else:
             if not file_defined:
                 lines.append(f'\t\t\t\t\t<file id="{master_id_to_use}">')
@@ -279,7 +308,13 @@ def generate_xml(sequence_name, segments, source_name, fps=25, start_offset_fram
         lines.append('\t\t\t\t\t<marker>')
         lines.append(f'\t\t\t\t\t\t<name>{clip_note}</name>')
         lines.append(f'\t\t\t\t\t\t<in>{src_in_frames}</in>')
-        lines.append(f'\t\t\t\t\t\t<out>{src_in_frames}</out>')
+        
+        # Extends the marker duration to span the entire clip for VO and Graphics
+        if is_vo or is_graphic:
+            lines.append(f'\t\t\t\t\t\t<out>{src_out_frames}</out>')
+        else:
+            lines.append(f'\t\t\t\t\t\t<out>{src_in_frames}</out>')
+            
         lines.append('\t\t\t\t\t</marker>')
 
         lines.append('\t\t\t\t</clipitem>')
@@ -296,8 +331,18 @@ def generate_xml(sequence_name, segments, source_name, fps=25, start_offset_fram
 
     timeline_head_frames = 0
     for i, seg in enumerate(segments, 1):
-        is_vo = seg.get('type') == 'vo' or 'text' in seg
+        seg_type = seg.get('type', 'clip')
+        is_vo = seg_type == 'vo' or ('text' in seg and seg_type != 'graphic')
+        is_graphic = seg_type == 'graphic'
         
+        if is_graphic:
+            # Graphics have no audio. We just advance the playhead so the timeline stays perfectly in sync!
+            duration_sec = float(seg.get('duration', 4.0))
+            duration_frames = seconds_to_frames(duration_sec, fps)
+            gap_frames = seconds_to_frames(float(seg.get('gap', 0.0)), fps)
+            timeline_head_frames += duration_frames + gap_frames
+            continue
+            
         if is_vo:
             vo_text = seg.get('text', 'VO')
             duration_sec = float(seg.get('duration', max(1.0, len(vo_text.split()) / 2.5)))
@@ -345,13 +390,20 @@ def generate_transcript_txt(sequence_name, segments, fps=25, start_offset_frames
     lines = [f"Sequence Transcript: {sequence_name}", "=" * 50, ""]
     
     for i, seg in enumerate(segments, 1):
-        is_vo = seg.get('type') == 'vo' or 'text' in seg
+        seg_type = seg.get('type', 'clip')
+        is_vo = seg_type == 'vo' or ('text' in seg and seg_type != 'graphic')
+        is_graphic = seg_type == 'graphic'
         
         if is_vo:
             vo_text = seg.get('text', 'VO')
             duration = float(seg.get('duration', max(1.0, len(vo_text.split()) / 2.5)))
             lines.append(f"Clip {i:02}  |  [GENERATED VOICE-OVER]  |  Est. Duration: {duration:.1f}s")
             lines.append(f"Script: \"{vo_text}\"")
+        elif is_graphic:
+            gfx_text = seg.get('text', 'Graphic Text')
+            duration = float(seg.get('duration', 4.0))
+            lines.append(f"Clip {i:02}  |  [GRAPHIC CARD]  |  Est. Duration: {duration:.1f}s")
+            lines.append(f"Text: \"{gfx_text}\"")
         else:
             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)))
@@ -388,9 +440,11 @@ def call_gemini_for_edl(transcript_data, story_prompt, api_key, previous_edit=No
     system_prompt = (
         "You are an expert Documentary Senior Editor. Use the provided transcript JSON "
         "(which includes Speaker IDs and word-level timestamps) to create a condensed story.\n\n"
-        "NEW CAPABILITY: You can now act as a Writer/Director and generate Voice-Over (VO) to bridge gaps, introduce topics, or summarize. "
-        "CRITICAL: ONLY generate 'vo' segments if the user explicitly requests voice over suggestions or narration in their creative brief. Otherwise, ONLY use 'clip' segments.\n\n"
-        "Output ONLY a valid JSON array of segments. Every segment MUST be either a 'clip' or a 'vo'.\n\n"
+        "NEW CAPABILITIES:\n"
+        "1. You can generate Voice-Over (VO) to bridge gaps, introduce topics, or summarize.\n"
+        "2. You can generate Graphic Cards (Title Cards) to display text on screen (e.g., location, date, or chapter title).\n"
+        "CRITICAL: ONLY generate 'vo' or 'graphic' segments if the user explicitly requests voice over, narration, or graphic/title cards in their creative brief. Otherwise, ONLY use 'clip' segments.\n\n"
+        "Output ONLY a valid JSON array of segments. Every segment MUST be a 'clip', a 'vo', or a 'graphic'.\n\n"
         "For 'clip' segments (extracting from the subject):\n"
         "{\"type\": \"clip\", \"src_start\": 12.5, \"src_end\": 25.0, \"note\": \"Subject talks about X\", \"gap\": 0.0}\n"
         "- IGNORE ALL INTERVIEWER COMMENTS.\n"
@@ -400,6 +454,8 @@ def call_gemini_for_edl(transcript_data, story_prompt, api_key, previous_edit=No
         "{\"type\": \"vo\", \"text\": \"The journey began years ago...\", \"duration\": 3.5, \"gap\": 0.0}\n"
         "- DURATION: Estimate duration accurately based on reading speed (approx 2.5 words per second).\n"
         "- Keep VO concise and in the style of a documentary narrator.\n\n"
+        "For 'graphic' segments (generating text on screen):\n"
+        "{\"type\": \"graphic\", \"text\": \"Chapter 1: The Beginning\", \"duration\": 4.0, \"gap\": 0.0}\n\n"
         "PACING: Group related clips. Add a 'gap' (in seconds) between distinct ideas."
     )
     
@@ -499,7 +555,7 @@ st.markdown("""
 * Upload your file here (video or audio).
 * 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.
-* **Note: If you want me to write voice over suggestions, please request it.**
+* **Note: If you want me to write voice over suggestions or title cards, please request it.**
 * 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.
 * **Resolve Users:** Uncheck "Automatically import source clips into media pool" during import.
@@ -641,7 +697,7 @@ if uploaded_file:
                 st.markdown(f"**{seg['speaker']}:** {seg['text']}")
         
         st.subheader("Your Instruction")
-        brief = st.text_area("What should the Junior Editor do?", placeholder="e.g. Find the most engaging bits. If you want me to write voice over suggestions, please request it.")
+        brief = st.text_area("What should the Junior Editor do?", placeholder="e.g. Find the most engaging bits. If you want me to write voice over suggestions or title cards, please request it.")
         
         if st.button("Generate Edit"):
             if not ACTIVE_GEMINI_KEY:
@@ -672,6 +728,8 @@ if uploaded_file:
                         
                         # Save to session state so buttons don't disappear on click
                         st.session_state.edit_generated = True
+                        st.session_state.revision_count = 1
+                        st.session_state.base_filename = safe_filename
                         st.session_state.final_output = final_output
                         st.session_state.ext = ext
                         st.session_state.transcript_txt = transcript_txt
@@ -694,7 +752,7 @@ if uploaded_file:
             
             st.divider()
             st.subheader("Revise Edit")
-            revision_brief = st.text_area("Want changes? Tell Junior Editor:", placeholder="e.g. Make it twice as long, add more details about X, or write a VO intro.")
+            revision_brief = st.text_area("Want changes? Tell Junior Editor:", placeholder="e.g. Make it twice as long, add a graphic card for location, or write a VO intro.")
             
             if st.button("Apply Revisions"):
                 with st.spinner("Junior Editor is revising the edit..."):
@@ -709,14 +767,18 @@ if uploaded_file:
                     )
                     
                     if new_edl_segments:
+                        # Increment version count
+                        st.session_state.revision_count = st.session_state.get("revision_count", 1) + 1
+                        rev_seq_name = f"{final_source_name} V{st.session_state.revision_count}"
+                        
                         if export_format == "EDL":
-                            final_output = generate_cmx_edl(final_source_name, new_edl_segments, final_source_name, fps, offset_frames)
+                            final_output = generate_cmx_edl(rev_seq_name, new_edl_segments, final_source_name, fps, offset_frames)
                             ext = "edl"
                         else:
-                            final_output = generate_xml(final_source_name, new_edl_segments, final_source_name, fps, offset_frames)
+                            final_output = generate_xml(rev_seq_name, new_edl_segments, final_source_name, fps, offset_frames)
                             ext = "xml"
                             
-                        transcript_txt = generate_transcript_txt(final_source_name, new_edl_segments, fps, offset_frames, st.session_state.transcript)
+                        transcript_txt = generate_transcript_txt(rev_seq_name, new_edl_segments, fps, offset_frames, st.session_state.transcript)
                         
                         st.session_state.final_output = final_output
                         st.session_state.ext = ext
@@ -724,6 +786,11 @@ if uploaded_file:
                         st.session_state.used_model = used_model
                         st.session_state.used_endpoint = used_endpoint
                         st.session_state.edl_segments = new_edl_segments
+                        
+                        # Update safe filename for the download buttons to show _V2, _V3, etc.
+                        st.session_state.safe_filename = f"{st.session_state.base_filename}_V{st.session_state.revision_count}"
+                        
                         st.rerun()
 
 
+