Update app.py

app.py

@@ -76,12 +76,11 @@ def save_frames(v, t, n):
 
 
 def clear_extracted_frames():
-    """Clear the selected frame sequence and reset trimming tied to that sequence."""
+    """Clear only the extracted-frame sequence; keep the independent trim setting unchanged."""
     return (
         [],
         None,
-        "✅ Extracted frames cleared. The final video will not have a frame intro.",
-        0,
+        "✅ Extracted frames cleared. The final video will not have a frame intro. Initial-frame trimming is unchanged.",
     )
 
 
@@ -164,11 +163,9 @@ def make_video(s, video_in, audio_in, trim_frames=0, speed_factor=1.0):
             return None, "Initial frames to remove must be 0 or greater."
         if not 0.5 <= speed_factor <= 1.5:
             return None, "Playback speed must be between 0.5x and 1.5x."
-        if trim_frames > 0 and not has_frames:
-            return None, "Extract frames first to use initial-frame trimming."
-
-        # Always use the currently selected input as the original-video segment. The extracted
-        # frame state is used only to build the intro, so those two parts stay independent.
+        # Always use the currently selected input as the original-video segment. Extracted-frame
+        # state is used only to build the optional intro. Source-video trimming works independently,
+        # whether or not any frames have been extracted.
         _, _, source_fps, source_duration = meta(v_path)
         trim_seconds = trim_frames / source_fps
         if trim_frames and trim_seconds >= source_duration:
@@ -178,27 +175,80 @@ def make_video(s, video_in, audio_in, trim_frames=0, speed_factor=1.0):
             )
 
         if not has_frames:
-            # With no intro frames, trim the source video if requested (the UI normally expects
-            # a frame intro for trimming, but this branch remains safe if called directly).
-            if trim_frames:
-                return None, "Extract frames first to use initial-frame trimming."
+            # No intro frames: create an output from the original video alone, applying any
+            # requested frame trim and/or replacement audio. This is intentionally independent
+            # of the extraction controls.
             base_out = v_path
-            msg = "✅ Final video created from the original video. No extracted-frame intro was added."
-            if audio_path:
-                base_out = os.path.join(tempfile.mkdtemp(), f"final_audio_{int(time.time() * 1000)}.mp4")
-                replacement_duration = audio_meta(audio_path)
-                loop_args = ["-stream_loop", "-1"] if replacement_duration > 0 and (source_duration - replacement_duration) > 1.0 else []
-                cmd = ["ffmpeg", "-y", "-i", v_path] + loop_args + [
-                    "-i", audio_path,
-                    "-map", "0:v:0", "-map", "1:a:0",
-                    "-t", str(source_duration),
-                    "-c:v", "copy", "-c:a", "aac", "-b:a", "192k",
-                    "-movflags", "+faststart", base_out,
-                ]
+            body_duration = max(0.0, source_duration - trim_seconds)
+            has_source_audio = stream_has_audio(v_path)
+            needs_processing = bool(trim_frames or audio_path)
+
+            if needs_processing:
+                base_out = os.path.join(
+                    tempfile.mkdtemp(), f"source_only_{int(time.time() * 1000)}.mp4"
+                )
+                video_filter = (
+                    f"[0:v:0]trim=start_frame={trim_frames},setpts=PTS-STARTPTS[vout]"
+                )
+                cmd = ["ffmpeg", "-y", "-i", v_path]
+
+                if audio_path:
+                    replacement_duration = audio_meta(audio_path)
+                    loop_args = (
+                        ["-stream_loop", "-1"]
+                        if replacement_duration > 0 and (body_duration - replacement_duration) > 1.0
+                        else []
+                    )
+                    cmd += loop_args + ["-i", audio_path]
+                    cmd += [
+                        "-filter_complex", video_filter,
+                        "-map", "[vout]", "-map", "1:a:0",
+                        "-t", str(body_duration),
+                        "-c:v", "libx264", "-crf", "18", "-preset", "medium",
+                        "-pix_fmt", "yuv420p", "-c:a", "aac", "-b:a", "192k",
+                        "-movflags", "+faststart", base_out,
+                    ]
+                elif has_source_audio:
+                    audio_filter = (
+                        f"{video_filter};[0:a:0]atrim=start={trim_seconds:.8f},"
+                        "asetpts=PTS-STARTPTS[aout]"
+                    )
+                    cmd += [
+                        "-filter_complex", audio_filter,
+                        "-map", "[vout]", "-map", "[aout]",
+                        "-t", str(body_duration),
+                        "-c:v", "libx264", "-crf", "18", "-preset", "medium",
+                        "-pix_fmt", "yuv420p", "-c:a", "aac", "-b:a", "192k",
+                        "-movflags", "+faststart", base_out,
+                    ]
+                else:
+                    cmd += [
+                        "-filter_complex", video_filter,
+                        "-map", "[vout]", "-an",
+                        "-t", str(body_duration),
+                        "-c:v", "libx264", "-crf", "18", "-preset", "medium",
+                        "-pix_fmt", "yuv420p", "-movflags", "+faststart", base_out,
+                    ]
+
                 result = subprocess.run(cmd, capture_output=True, text=True)
                 if result.returncode:
-                    return None, result.stderr or "Could not replace the audio."
-                msg = "✅ Final video created using the original video and replacement audio."
+                    return None, result.stderr or "Could not process the original video."
+
+            if trim_frames:
+                msg = (
+                    "✅ Final video created from the original video with no extracted-frame intro. "
+                    f"Skipped the first {trim_frames} frame(s) ({trim_seconds:.3f}s at "
+                    f"{source_fps:.3f} FPS)."
+                )
+            else:
+                msg = "✅ Final video created from the original video. No extracted-frame intro was added."
+            if audio_path:
+                msg = "✅ Final video created from the original video with replacement audio and no extracted-frame intro."
+            elif has_source_audio and trim_frames:
+                msg += " Original audio was trimmed to stay aligned with the video."
+            elif not has_source_audio:
+                msg += " The source video has no audio track."
+
             final_out = apply_speed_to_video(base_out, speed_factor)
             return final_out, add_speed_message(msg, speed_factor)
 
@@ -345,7 +395,7 @@ with gr.Blocks(title="Video Frame Editor") as demo:
     clear_btn.click(
         fn=clear_extracted_frames,
         inputs=[],
-        outputs=[gallery, state, status, trim_count],
+        outputs=[gallery, state, status],
     )
     copy_btn.click(
         fn=make_video,