able to re-transcribe

backend/routers/export.py

@@ -55,6 +55,13 @@ class ExportRequest(BaseModel):
     deleted_indices: Optional[List[int]] = None
 
 
+class TranscriptExportRequest(BaseModel):
+    words: List[ExportWordModel]
+    deleted_indices: Optional[List[int]] = None
+    output_path: str
+    format: str = "txt"  # "txt" or "srt"
+
+
 def _map_ranges_to_output_timeline(
     ranges: List[dict],
     keep_segments: List[dict],
@@ -234,3 +241,34 @@ async def export_video(req: ExportRequest):
     except Exception as e:
         logger.error(f"Export error: {e}", exc_info=True)
         raise HTTPException(status_code=500, detail=str(e))
+
+
+@router.post("/export/transcript")
+async def export_transcript(req: TranscriptExportRequest):
+    """Export transcript as plain text or SRT without rendering video."""
+    try:
+        from services.caption_generator import generate_srt
+
+        deleted_set = set(req.deleted_indices or [])
+        word_dicts = [w.model_dump() for w in req.words]
+
+        if req.format == "srt":
+            content = generate_srt(word_dicts, deleted_set)
+        else:
+            # Plain text: join non-deleted words
+            active_words = []
+            for i, w in enumerate(word_dicts):
+                if i not in deleted_set:
+                    active_words.append(w["word"])
+            content = " ".join(active_words)
+
+        os.makedirs(os.path.dirname(req.output_path) or ".", exist_ok=True)
+        with open(req.output_path, "w", encoding="utf-8") as f:
+            f.write(content)
+
+        logger.info("Transcript exported to %s (format=%s)", req.output_path, req.format)
+        return {"status": "ok", "output_path": req.output_path}
+
+    except Exception as e:
+        logger.error(f"Transcript export failed: {e}", exc_info=True)
+        raise HTTPException(status_code=500, detail=str(e))

backend/routers/transcribe.py

@@ -51,3 +51,99 @@ async def transcribe(req: TranscribeRequest):
     except Exception as e:
         logger.error(f"Transcription failed: {e}", exc_info=True)
         raise HTTPException(status_code=500, detail=str(e))
+
+
+class ReTranscribeSegmentRequest(BaseModel):
+    file_path: str
+    start: float
+    end: float
+    model: str = "base"
+    language: Optional[str] = None
+
+
+@router.post("/transcribe/segment")
+async def transcribe_segment(req: ReTranscribeSegmentRequest):
+    """
+    Re-transcribe a specific segment of audio.
+    Extracts the segment with FFmpeg, transcribes it, and returns words
+    with timestamps adjusted to the original file timeline.
+    """
+    import subprocess
+    import tempfile
+    import os
+
+    try:
+        # Extract the segment to a temp file
+        tmp_dir = tempfile.mkdtemp(prefix="talkedit_segment_")
+        segment_path = os.path.join(tmp_dir, "segment.wav")
+
+        cmd = [
+            "ffmpeg", "-y",
+            "-i", req.file_path,
+            "-ss", str(req.start),
+            "-to", str(req.end),
+            "-vn",
+            "-acodec", "pcm_s16le",
+            "-ar", "16000",
+            "-ac", "1",
+            segment_path,
+        ]
+        result = subprocess.run(cmd, capture_output=True, text=True)
+        if result.returncode != 0:
+            raise RuntimeError(f"Segment extraction failed: {result.stderr[-300:]}")
+
+        # Transcribe the segment — try GPU first, fall back to CPU
+        try:
+            segment_result = transcribe_audio(
+                file_path=segment_path,
+                model_name=req.model,
+                use_gpu=True,
+                use_cache=False,
+                language=req.language,
+            )
+        except Exception as gpu_err:
+            logger.warning(f"GPU transcription failed (%s), falling back to CPU", gpu_err)
+            segment_result = transcribe_audio(
+                file_path=segment_path,
+                model_name=req.model,
+                use_gpu=False,
+                use_cache=False,
+                language=req.language,
+            )
+
+        # Adjust timestamps to be relative to the original file
+        offset = req.start
+        adjusted_words = []
+        for w in segment_result.get("words", []):
+            w["start"] = round(w["start"] + offset, 3)
+            w["end"] = round(w["end"] + offset, 3)
+            adjusted_words.append(w)
+
+        adjusted_segments = []
+        for seg in segment_result.get("segments", []):
+            seg["start"] = round(seg["start"] + offset, 3)
+            seg["end"] = round(seg["end"] + offset, 3)
+            # Also adjust words within each segment
+            for w in seg.get("words", []):
+                w["start"] = round(w["start"] + offset, 3)
+                w["end"] = round(w["end"] + offset, 3)
+            adjusted_segments.append(seg)
+
+        # Cleanup
+        try:
+            os.remove(segment_path)
+            os.rmdir(tmp_dir)
+        except OSError:
+            pass
+
+        return {
+            "words": adjusted_words,
+            "segments": adjusted_segments,
+            "language": segment_result.get("language", "en"),
+        }
+
+    except FileNotFoundError:
+        raise HTTPException(status_code=404, detail=f"File not found: {req.file_path}")
+    except Exception as e:
+        logger.error(f"Segment transcription failed: {e}", exc_info=True)
+        raise HTTPException(status_code=500, detail=str(e))