Fix SRT subtitles: word-level clipping for partial segments

Bug found in Žena ME TEPE re-test:
- Clip start: 76.73s (correct, captures full 'Žena' word)
- But SRT subtitle #1 showed: 'SAJ ŠE DOMA MI VEČ NOČJO VERJET.'
- That text is from the PREVIOUS verse, not the chorus!

Why: previous segment (73.9-78.2s) contained 'saj še doma mi več
nočjo verjet. Žena me'. Clip start fell at 76.73s (mid-segment).
Old SRT logic: max(s_start, clip_start) just clipped TIMING but kept
ALL the text from that segment, including text from before the clip.

Fix: when a segment partially falls outside clip range AND has word-level
timestamps (Scribe provides these), reconstruct the segment using only
the words that actually fall within [clip_start, clip_end]. Audio
(clipped at clip_start) only contains those words anyway, so the
subtitle should match.

Result for Žena chorus:
- Old: 'SAJ ŠE DOMA MI VEČ NOČJO VERJET.' (wrong, that text is silent
        in clip)
- New: 'ŽENA ME' (only words actually heard at 76.73-78.16s)

app/main.py

@@ -224,13 +224,43 @@ def generate_srt_from_segments(segments, clip_start, clip_end, output_path):
         s_start = float(seg["start"])
         s_end = float(seg["end"])
         text = str(seg["text"]).strip()
+        words = seg.get("words", []) or []
 
         # Filter v range
         if s_end <= clip_start or s_start >= clip_end:
             continue
-        # Klipni
-        s_start = max(s_start, clip_start)
-        s_end = min(s_end, clip_end)
+        
+        # Če segment delno štrli iz clip range-a IN imamo word-level timestampe,
+        # uporabi samo tiste besede ki dejansko padejo v clip range
+        # (sicer subtitle vsebuje besedilo iz prejšnjega/naslednjega refrena/verza)
+        if words and (s_start < clip_start or s_end > clip_end):
+            words_in_clip = []
+            for w in words:
+                w_start = float(w.get("start", 0))
+                w_end = float(w.get("end", 0))
+                w_text = w.get("text", "").strip()
+                if not w_text:
+                    continue
+                # Beseda padeva v clip če se prekriva (ne mora biti popolnoma znotraj)
+                if w_end > clip_start and w_start < clip_end:
+                    words_in_clip.append({
+                        "start": max(w_start, clip_start),
+                        "end": min(w_end, clip_end),
+                        "text": w_text,
+                    })
+            
+            if not words_in_clip:
+                continue
+            
+            # Reconstruiraj segment z dejanskim word-level timing-om
+            text = " ".join(w["text"] for w in words_in_clip)
+            s_start = words_in_clip[0]["start"]
+            s_end = words_in_clip[-1]["end"]
+        else:
+            # Klipni segment začetek/konec na clip range
+            s_start = max(s_start, clip_start)
+            s_end = min(s_end, clip_end)
+        
         if s_end - s_start < 0.2:
             continue