fixed dropdown bar visibility

.github/copilot-instructions.md

@@ -55,6 +55,16 @@ Use project virtualenvs where available (`.venv312`, `.venv`, or `venv`) for bac
 - Media URL handling between project load paths should remain consistent to avoid format-specific regressions (especially WAV/MP3 behavior).
 - Export pipeline changes must preserve caption modes (`none`, `sidecar`, `burn-in`) and audio enhancement behavior.
 
+## Recent Changes
+
+### 2026-05-04 — Word text correction, low-confidence highlighting, audio normalization
+
+- **Word text correction (#015)**: Double-click any word in the transcript editor to edit its text inline. Press Enter to commit, Escape to cancel. State is updated in both `words[]` and `segments[]` arrays (segment text recomposed from updated words). Pure frontend; no backend changes needed.
+- **Low-confidence word highlighting (#012)**: Words with `confidence < threshold` (default 0.6, configurable in Settings panel) render with an orange dotted underline. Tooltip shows exact confidence percentage. Threshold is persisted in `localStorage` key `talkedit:confidenceThreshold`.
+- **Audio normalization (#018)**: New backend endpoint `POST /audio/normalize` in `backend/routers/audio.py`. Two-pass FFmpeg `loudnorm` (measure then apply) implemented in `backend/services/audio_cleaner.py:normalize_audio()`. Falls back to single-pass if measurement fails. Frontend UI in Export panel: target selector (YouTube -14, Spotify -16, Broadcast -23, etc.) with "Normalize" button.
+- **Store**: New `updateWordText(index, text)` action in `editorStore.ts` updates both `words[]` and recomputes `segments[].text`.
+- **Settings panel**: New confidence threshold slider (0–1 range).
+
 ## Update Rules (Important)
 
 When a task changes architecture, app wiring, commands, API shape, project schema, or major conventions, update this file before finishing.

FEATURES.md

@@ -6,13 +6,13 @@ Features are grouped by priority. Check off items as they are implemented.
 
 ## 🔴 Highest Impact Next — Conversion and retention
 
-- [ ] [#015] **Word text correction** — allow editing the transcript text of a word without affecting its timing. Whisper gets homophones/proper nouns wrong constantly. Pure frontend state change; no backend needed.
+- [x] [#015] **Word text correction** — double-click any word to edit its text in-place. Preserves timing and confidence. Pure frontend state change. (2026-05-04)
 
 - [ ] [#013] **Re-transcribe selection** — if Whisper gets a section wrong, let the user select a word range and re-run transcription on just that segment (optionally with a different model or language).
 
-- [ ] [#012] **Low-confidence word highlighting** — WhisperX already returns `confidence` per word. Words below a threshold (e.g. < 0.6) should be visually underlined or tinted so the user knows where to double-check.
+- [x] [#012] **Low-confidence word highlighting** — words with `confidence < 0.6` (configurable in Settings) get an orange dotted underline. Hover shows exact confidence %. (2026-05-04)
 
-- [ ] [#018] **Audio normalization / loudness targeting** — single "Normalize" button that targets a LUFS level (-14 for YouTube, -16 for Spotify). Backend: `ffmpeg -af loudnorm`. Very high value for podcasters, ~2–3 hours of work.
+- [x] [#018] **Audio normalization / loudness targeting** — "Normalize" button in Export panel with LUFS target selector (-14 YouTube, -16 Spotify, -23 Broadcast). Backend: FFmpeg two-pass `loudnorm`. (2026-05-04)
 
 - [ ] [#024] **Export to transcript text / SRT only** — some users just want a clean `.txt` or `.srt` of the edited transcript without rendering video.
 

backend/routers/audio.py

@@ -11,7 +11,7 @@ from fastapi import APIRouter, HTTPException, Query, Request
 from fastapi.responses import FileResponse
 from pydantic import BaseModel
 
-from services.audio_cleaner import clean_audio, detect_silence_ranges, is_deepfilter_available
+from services.audio_cleaner import clean_audio, detect_silence_ranges, is_deepfilter_available, normalize_audio
 
 logger = logging.getLogger(__name__)
 router = APIRouter()
@@ -164,3 +164,30 @@ async def get_waveform_audio(request: Request, path: str = Query(...)):
     )
     _waveform_cache[cache_key] = str(out_wav)
     return FileResponse(str(out_wav), media_type="audio/wav")
+
+
+class NormalizeRequest(BaseModel):
+    input_path: str
+    output_path: Optional[str] = None
+    target_lufs: float = -14.0
+
+
+@router.post("/audio/normalize")
+async def normalize_audio_endpoint(req: NormalizeRequest):
+    """Normalize audio loudness to a target LUFS level using FFmpeg loudnorm."""
+    if req.target_lufs < -70 or req.target_lufs > 0:
+        raise HTTPException(status_code=400, detail="target_lufs must be between -70 and 0")
+    try:
+        output = normalize_audio(
+            req.input_path,
+            req.output_path or "",
+            target_lufs=req.target_lufs,
+        )
+        return {
+            "status": "ok",
+            "output_path": output,
+            "target_lufs": req.target_lufs,
+        }
+    except Exception as e:
+        logger.error(f"Audio normalization failed: {e}", exc_info=True)
+        raise HTTPException(status_code=500, detail=str(e))

backend/routers/export.py

@@ -48,6 +48,8 @@ class ExportRequest(BaseModel):
     resolution: str = "1080p"
     format: str = "mp4"
     enhanceAudio: bool = False
+    normalize_loudness: bool = False
+    normalize_target_lufs: float = -14.0
     captions: str = "none"
     words: Optional[List[ExportWordModel]] = None
     deleted_indices: Optional[List[int]] = None
@@ -166,6 +168,8 @@ async def export_video(req: ExportRequest):
                     gain_ranges=mapped_gain_segments,
                     speed_ranges=speed_segments,
                     global_gain_db=req.global_gain_db,
+                    normalize_loudness=req.normalize_loudness,
+                    normalize_target_lufs=req.normalize_target_lufs,
                 )
             else:
                 output = export_reencode(
@@ -178,6 +182,8 @@ async def export_video(req: ExportRequest):
                     gain_ranges=mapped_gain_segments,
                     speed_ranges=speed_segments,
                     global_gain_db=req.global_gain_db,
+                    normalize_loudness=req.normalize_loudness,
+                    normalize_target_lufs=req.normalize_target_lufs,
                 )
         finally:
             if ass_path and os.path.exists(ass_path):

backend/services/audio_cleaner.py

@@ -158,3 +158,125 @@ def detect_silence_ranges(input_path: str, min_silence_ms: int, silence_db: floa
         silence_db,
     )
     return ranges
+
+
+def normalize_audio(
+    input_path: str,
+    output_path: str = "",
+    target_lufs: float = -14.0,
+) -> str:
+    """
+    Normalize audio loudness to a target LUFS level using FFmpeg's loudnorm filter.
+
+    Args:
+        input_path: Path to the input audio/video file.
+        output_path: Path for the normalized output. Auto-generated if empty.
+        target_lufs: Target integrated loudness in LUFS.
+                     Common targets: -14 (YouTube), -16 (Spotify), -23 (broadcast).
+
+    Returns: path to the normalized audio file.
+    """
+    import os as _os
+
+    inp = Path(input_path)
+    if not output_path:
+        output_path = str(inp.with_stem(inp.stem + "_normalized"))
+
+    # Two-pass loudnorm: first pass measures loudness, second pass applies correction.
+    # First pass: measure only (print_format=json)
+    measure_cmd = [
+        "ffmpeg", "-y",
+        "-i", str(inp),
+        "-af", f"loudnorm=I={target_lufs}:LRA=7:TP=-1.5:print_format=json",
+        "-f", "null",
+        "-",
+    ]
+    logger.info("Running loudnorm first pass (measurement): %s", " ".join(measure_cmd))
+    measure_result = subprocess.run(measure_cmd, capture_output=True, text=True)
+
+    # Parse measured parameters from stderr (loudnorm outputs JSON to stderr)
+    measured = _parse_loudnorm_measurement(measure_result.stderr)
+    if not measured:
+        logger.warning(
+            "loudnorm measurement failed or produced no output; "
+            "falling back to single-pass normalization"
+        )
+        # Single-pass fallback
+        cmd = [
+            "ffmpeg", "-y",
+            "-i", str(inp),
+            "-af", f"loudnorm=I={target_lufs}:LRA=7:TP=-1.5",
+            "-c:v", "copy",
+            output_path,
+        ]
+        result = subprocess.run(cmd, capture_output=True, text=True)
+        if result.returncode != 0:
+            raise RuntimeError(f"Audio normalization failed: {result.stderr[-300:]}")
+        logger.info("Single-pass normalized audio saved to %s", output_path)
+        return output_path
+
+    # Second pass: apply normalization using measured values
+    input_i = measured.get("input_i", target_lufs)
+    input_lra = measured.get("input_lra", 7.0)
+    input_tp = measured.get("input_tp", -1.5)
+    input_thresh = measured.get("input_thresh", -30.0)
+    offset = measured.get("target_offset", 0.0)
+
+    apply_cmd = [
+        "ffmpeg", "-y",
+        "-i", str(inp),
+        "-af",
+        (
+            f"loudnorm=I={target_lufs}:LRA=7:TP=-1.5:"
+            f"measured_I={input_i}:measured_LRA={input_lra}:"
+            f"measured_TP={input_tp}:measured_thresh={input_thresh}:"
+            f"offset={offset}:linear=true:print_format=summary"
+        ),
+        "-c:v", "copy",
+        output_path,
+    ]
+    logger.info("Running loudnorm second pass (apply): %s", " ".join(apply_cmd))
+    result = subprocess.run(apply_cmd, capture_output=True, text=True)
+    if result.returncode != 0:
+        raise RuntimeError(f"Audio normalization (apply) failed: {result.stderr[-300:]}")
+
+    logger.info(
+        "Normalized audio saved to %s (target=%s LUFS, measured_I=%s)",
+        output_path,
+        target_lufs,
+        input_i,
+    )
+    return output_path
+
+
+def _parse_loudnorm_measurement(stderr_output: str) -> dict:
+    """Parse loudnorm JSON measurement output from FFmpeg stderr."""
+    import json
+
+    # loudnorm prints JSON block between "Parsed_loudnorm" lines
+    lines = stderr_output.split("\n")
+    json_lines = []
+    in_json = False
+    for line in lines:
+        if "Parsed_loudnorm" in line and "}" in line:
+            # Single-line JSON
+            try:
+                start = line.index("{")
+                end = line.rindex("}") + 1
+                return json.loads(line[start:end])
+            except (ValueError, json.JSONDecodeError):
+                continue
+        if "{" in line and not in_json:
+            in_json = True
+        if in_json:
+            json_lines.append(line)
+        if in_json and "}" in line:
+            in_json = False
+            break
+
+    if json_lines:
+        try:
+            return json.loads("".join(json_lines))
+        except json.JSONDecodeError:
+            pass
+    return {}

backend/services/video_editor.py

@@ -13,6 +13,20 @@ from typing import List
 logger = logging.getLogger(__name__)
 
 
+def _get_codec_args(format_hint: str, has_video: bool = True) -> list:
+    """Return FFmpeg codec arguments for the given format."""
+    if format_hint == "wav":
+        return ["-c:a", "pcm_s16le"]
+    if format_hint == "webm":
+        if has_video:
+            return ["-c:v", "libvpx-vp9", "-crf", "30", "-b:v", "0", "-c:a", "libopus"]
+        return ["-c:a", "libopus", "-b:a", "160k"]
+    # Default: MP4
+    if has_video:
+        return ["-c:v", "libx264", "-preset", "medium", "-crf", "18", "-c:a", "aac", "-b:a", "192k"]
+    return ["-c:a", "aac", "-b:a", "192k"]
+
+
 def _input_has_video_stream(ffmpeg_cmd: str, input_path: str) -> bool:
     """Return True if the input contains at least one video stream."""
     ffprobe = ffmpeg_cmd.replace("ffmpeg", "ffprobe")
@@ -135,7 +149,7 @@ def export_stream_copy(
         output_path on success
     """
     if mute_ranges:
-        # Mute ranges require audio filtering, so fall back to re-encoding
+        # Mute ranges require audio filtering, so fall back to re-encode
         return export_reencode(input_path, output_path, keep_segments, "1080p", "mp4", mute_ranges)
     ffmpeg = _find_ffmpeg()
     if not _input_has_video_stream(ffmpeg, input_path):
@@ -209,6 +223,8 @@ def export_reencode(
     gain_ranges: List[dict] = None,
     speed_ranges: List[dict] = None,
     global_gain_db: float = 0.0,
+    normalize_loudness: bool = False,
+    normalize_target_lufs: float = -14.0,
 ) -> str:
     """
     Export video with full re-encode. Slower but supports resolution changes,
@@ -241,6 +257,9 @@ def export_reencode(
             end = mute_range['end']
             filters.append(f"volume=0:enable='between(t,{start},{end})'")
 
+        if normalize_loudness:
+            filters.append(f"loudnorm=I={normalize_target_lufs}:LRA=7:TP=-1.5")
+
         return ",".join(filters) if filters else "anull"
 
     has_audio_filters = bool(mute_ranges) or bool(gain_ranges) or abs(float(global_gain_db)) > 1e-6
@@ -278,16 +297,14 @@ def export_reencode(
 
         filter_complex = "".join(filter_parts)
 
-        audio_codec_args = ["-c:a", "aac", "-b:a", "192k"]
+        codec_args = _get_codec_args(format_hint, has_video=False)
-        if format_hint == "webm":
-            audio_codec_args = ["-c:a", "libopus", "-b:a", "160k"]
 
         cmd = [
             ffmpeg, "-y",
             "-i", input_path,
             "-filter_complex", filter_complex,
             "-map", audio_map,
-            *audio_codec_args,
+            *codec_args,
             output_path,
         ]
 
@@ -319,9 +336,7 @@ def export_reencode(
 
         filter_complex = f"[0:a]{audio_filter}[a];[0:v]{video_filter}{video_map}"
 
-        codec_args = ["-c:v", "libx264", "-preset", "medium", "-crf", "18", "-c:a", "aac", "-b:a", "192k"]
+        codec_args = _get_codec_args(format_hint, has_video)
-        if format_hint == "webm":
-            codec_args = ["-c:v", "libvpx-vp9", "-crf", "30", "-b:v", "0", "-c:a", "libopus"]
 
         cmd = [
             ffmpeg, "-y",
@@ -367,6 +382,12 @@ def export_reencode(
 
         filter_complex = "".join(filter_parts)
 
+        # Add loudnorm to the cutting path audio chain if enabled
+        audio_map_label = "[outa]"
+        if normalize_loudness:
+            filter_complex += f";{audio_map_label}loudnorm=I={normalize_target_lufs}:LRA=7:TP=-1.5[outa_norm]"
+            audio_map_label = "[outa_norm]"
+
         scale = scale_map.get(resolution, "")
         if scale:
             filter_complex += f";[outv]{scale}[outv_scaled]"
@@ -374,25 +395,24 @@ def export_reencode(
         else:
             video_map = "[outv]"
 
-        codec_args = ["-c:v", "libx264", "-preset", "medium", "-crf", "18", "-c:a", "aac", "-b:a", "192k"]
+        codec_args = _get_codec_args(format_hint, has_video)
-        if format_hint == "webm":
-            codec_args = ["-c:v", "libvpx-vp9", "-crf", "30", "-b:v", "0", "-c:a", "libopus"]
 
         cmd = [
             ffmpeg, "-y",
             "-i", input_path,
             "-filter_complex", filter_complex,
             "-map", video_map,
-            "-map", "[outa]",
+            "-map", audio_map_label,
             *codec_args,
             "-movflags", "+faststart",
             output_path,
         ]
 
         logger.info(
-            "Re-encoding %s segments (speed-adjusted=%s) -> %s (%s)",
+            "Re-encoding %s segments (speed-adjusted=%s, normalize=%s) -> %s (%s)",
             n,
             has_speed,
+            normalize_loudness,
             output_path,
             resolution,
         )
@@ -415,6 +435,8 @@ def export_reencode_with_subs(
     gain_ranges: List[dict] = None,
     speed_ranges: List[dict] = None,
     global_gain_db: float = 0.0,
+    normalize_loudness: bool = False,
+    normalize_target_lufs: float = -14.0,
 ) -> str:
     """
     Export video with re-encode and burn-in subtitles (ASS format).
@@ -451,6 +473,9 @@ def export_reencode_with_subs(
             end = mute_range['end']
             filters.append(f"volume=0:enable='between(t,{start},{end})'")
 
+        if normalize_loudness:
+            filters.append(f"loudnorm=I={normalize_target_lufs}:LRA=7:TP=-1.5")
+
         return ",".join(filters) if filters else "anull"
 
     has_audio_filters = bool(mute_ranges) or bool(gain_ranges) or abs(float(global_gain_db)) > 1e-6
@@ -472,9 +497,7 @@ def export_reencode_with_subs(
 
         filter_complex = f"[0:a]{audio_filter}[a];[0:v]{video_filter}[v]"
 
-        codec_args = ["-c:v", "libx264", "-preset", "medium", "-crf", "18", "-c:a", "aac", "-b:a", "192k"]
+        codec_args = _get_codec_args(format_hint, has_video=True)
-        if format_hint == "webm":
-            codec_args = ["-c:v", "libvpx-vp9", "-crf", "30", "-b:v", "0", "-c:a", "libopus"]
 
         cmd = [
             ffmpeg, "-y",
@@ -530,9 +553,7 @@ def export_reencode_with_subs(
             filter_complex += f";[outv]ass='{escaped_sub}'[outv_final]"
         video_map = "[outv_final]"
 
-        codec_args = ["-c:v", "libx264", "-preset", "medium", "-crf", "18", "-c:a", "aac", "-b:a", "192k"]
+        codec_args = _get_codec_args(format_hint, has_video=True)
-        if format_hint == "webm":
-            codec_args = ["-c:v", "libvpx-vp9", "-crf", "30", "-b:v", "0", "-c:a", "libopus"]
 
         cmd = [
             ffmpeg, "-y",

frontend/src/components/ExportDialog.tsx

@@ -1,6 +1,6 @@
 import { useState, useCallback } from 'react';
 import { useEditorStore } from '../store/editorStore';
-import { Download, Loader2, Zap, Cog, Info } from 'lucide-react';
+import { Download, Loader2, Zap, Cog, Info, Volume2 } from 'lucide-react';
 import type { ExportOptions } from '../types/project';
 
 export default function ExportDialog() {
@@ -9,25 +9,40 @@ export default function ExportDialog() {
 
   const hasCuts = cutRanges.length > 0;
 
-  const [options, setOptions] = useState<Omit<ExportOptions, 'outputPath'>>({
+  // Detect if input is audio-only by its extension
-    mode: 'fast',
+  const audioExtensions = new Set(['.wav', '.mp3', '.flac', '.m4a', '.ogg', '.aac', '.wma']);
+  const inputExt = videoPath ? '.' + videoPath.split('.').pop()?.toLowerCase() : '';
+  const isAudioOnly = videoPath ? audioExtensions.has(inputExt) : false;
+
+  const [options, setOptions] = useState<Omit<ExportOptions, 'outputPath'> & { normalizeAudio: boolean; normalizeTarget: number }>({
+    mode: isAudioOnly ? 'reencode' : 'fast',
     resolution: '1080p',
-    format: 'mp4',
+    format: isAudioOnly ? 'wav' : 'mp4',
     enhanceAudio: false,
     captions: 'none',
+    normalizeAudio: false,
+    normalizeTarget: -14,
   });
   const [exportError, setExportError] = useState<string | null>(null);
 
+  const HANDLE_EXPORT_filters = useCallback(() => {
+    const ext = options.format;
+    const nameMap: Record<string, string> = {
+      mp4: 'MP4',
+      mov: 'MOV',
+      webm: 'WebM',
+      wav: 'WAV Audio',
+    };
+    return [{ name: nameMap[ext] || 'File', extensions: [ext] }];
+  }, [options.format]);
+
   const handleExport = useCallback(async () => {
     if (!videoPath) return;
 
+    const defaultExt = options.format === 'wav' ? 'wav' : 'mp4';
     const outputPath = await window.electronAPI?.saveFile({
-      defaultPath: videoPath.replace(/\.[^.]+$/, '_edited.mp4'),
+      defaultPath: videoPath.replace(/\.[^.]+$/, `_edited.${defaultExt}`),
-      filters: [
+      filters: HANDLE_EXPORT_filters(),
-        { name: 'MP4', extensions: ['mp4'] },
-        { name: 'MOV', extensions: ['mov'] },
-        { name: 'WebM', extensions: ['webm'] },
-      ],
     });
     if (!outputPath) return;
 
@@ -44,6 +59,18 @@ export default function ExportDialog() {
         }
       }
 
+      // Map frontend camelCase gain/speed fields to backend snake_case
+      const backendGainRanges = gainRanges.map((r) => ({
+        start: r.start,
+        end: r.end,
+        gain_db: r.gainDb,
+      }));
+      const backendSpeedRanges = speedRanges.map((r) => ({
+        start: r.start,
+        end: r.end,
+        speed: r.speed,
+      }));
+
       const res = await fetch(`${backendUrl}/export`, {
         method: 'POST',
         headers: { 'Content-Type': 'application/json' },
@@ -51,13 +78,19 @@ export default function ExportDialog() {
           input_path: videoPath,
           output_path: outputPath,
           keep_segments: keepSegments,
-          mute_ranges: muteRanges,
+          mute_ranges: muteRanges.length > 0 ? muteRanges.map((r) => ({ start: r.start, end: r.end })) : undefined,
-          gain_ranges: gainRanges,
+          gain_ranges: backendGainRanges.length > 0 ? backendGainRanges : undefined,
-          speed_ranges: speedRanges,
+          speed_ranges: backendSpeedRanges.length > 0 ? backendSpeedRanges : undefined,
           global_gain_db: globalGainDb,
           words: options.captions !== 'none' ? words : undefined,
           deleted_indices: options.captions !== 'none' ? [...deletedSet] : undefined,
-          ...options,
+          mode: options.mode,
+          resolution: options.resolution,
+          format: options.format,
+          enhanceAudio: options.enhanceAudio,
+          normalize_loudness: options.normalizeAudio,
+          normalize_target_lufs: options.normalizeTarget,
+          captions: options.captions,
         }),
       });
       if (!res.ok) {
@@ -76,7 +109,7 @@ export default function ExportDialog() {
       setExportError(err instanceof Error ? err.message : 'Export failed');
       setExporting(false);
     }
-  }, [videoPath, options, backendUrl, setExporting, getKeepSegments, cutRanges, muteRanges, gainRanges, speedRanges, globalGainDb, words]);
+  }, [videoPath, options, backendUrl, setExporting, getKeepSegments, cutRanges, muteRanges, gainRanges, speedRanges, globalGainDb, words, HANDLE_EXPORT_filters]);
 
   return (
     <div className="p-4 space-y-5">
@@ -126,9 +159,44 @@ export default function ExportDialog() {
           { value: 'mp4', label: 'MP4 (H.264)' },
           { value: 'mov', label: 'MOV (QuickTime)' },
           { value: 'webm', label: 'WebM (VP9)' },
+          ...(isAudioOnly ? [{ value: 'wav' as const, label: 'WAV (Uncompressed)' }] : []),
         ]}
       />
 
+      {/* Audio normalization — integrated into export */}
+      <div className="space-y-2 pt-1 border-t border-editor-border">
+        <label className="flex items-center gap-2 cursor-pointer">
+          <input
+            type="checkbox"
+            checked={options.normalizeAudio}
+            onChange={(e) => setOptions((o) => ({ ...o, normalizeAudio: e.target.checked }))}
+            className="w-4 h-4 rounded bg-editor-surface border-editor-border accent-editor-accent"
+          />
+          <div>
+            <span className="text-xs font-medium">Normalize loudness</span>
+            <p className="text-[10px] text-editor-text-muted">
+              Apply LUFS normalization during export. Requires re-encode.
+            </p>
+          </div>
+        </label>
+        {options.normalizeAudio && (
+          <div className="flex items-center gap-2 pl-6">
+            <Volume2 className="w-3 h-3 text-editor-text-muted shrink-0" />
+            <select
+              value={options.normalizeTarget}
+              onChange={(e) => setOptions((o) => ({ ...o, normalizeTarget: Number(e.target.value) }))}
+              className="flex-1 px-2 py-1.5 text-xs bg-editor-surface border border-editor-border rounded focus:outline-none focus:border-editor-accent [color-scheme:dark]"
+            >
+              <option value={-14}>YouTube (-14 LUFS)</option>
+              <option value={-16}>Spotify (-16 LUFS)</option>
+              <option value={-23}>Broadcast (-23 LUFS)</option>
+              <option value={-11}>Loud (-11 LUFS)</option>
+              <option value={-9}>Very Loud (-9 LUFS)</option>
+            </select>
+          </div>
+        )}
+      </div>
+
       {/* Audio enhancement */}
       <label className="flex items-center gap-2 cursor-pointer">
         <input
@@ -241,7 +309,7 @@ function SelectField({
       <select
         value={value}
         onChange={(e) => onChange(e.target.value)}
-        className="w-full px-3 py-2 bg-editor-surface border border-editor-border rounded-lg text-xs text-editor-text focus:outline-none focus:border-editor-accent"
+        className="w-full px-3 py-2 bg-editor-surface border border-editor-border rounded-lg text-xs text-editor-text focus:outline-none focus:border-editor-accent [color-scheme:dark]"
       >
         {options.map((opt) => (
           <option key={opt.value} value={opt.value}>

frontend/src/components/SettingsPanel.tsx

@@ -7,6 +7,18 @@ import { Bot, Cloud, Brain, RefreshCw } from 'lucide-react';
 export default function SettingsPanel() {
   const { providers, defaultProvider, setProviderConfig, setDefaultProvider } = useAIStore();
   const { backendUrl, zonePreviewPaddingSeconds, setZonePreviewPaddingSeconds } = useEditorStore();
+  const CONFIDENCE_THRESHOLD_KEY = 'talkedit:confidenceThreshold';
+  const [confidenceThreshold, setConfidenceThresholdState] = useState(() => {
+    const stored = typeof window !== 'undefined' ? Number(window.localStorage.getItem(CONFIDENCE_THRESHOLD_KEY)) : 0;
+    return Number.isFinite(stored) ? stored : 0.6;
+  });
+  const setConfidenceThreshold = (value: number) => {
+    const clamped = Math.max(0, Math.min(1, value));
+    setConfidenceThresholdState(clamped);
+    if (typeof window !== 'undefined') {
+      window.localStorage.setItem(CONFIDENCE_THRESHOLD_KEY, String(clamped));
+    }
+  };
   const [ollamaModels, setOllamaModels] = useState<string[]>([]);
   const [loadingModels, setLoadingModels] = useState(false);
 
@@ -66,6 +78,40 @@ export default function SettingsPanel() {
         </div>
       </ProviderSection>
 
+      {/* Confidence threshold */}
+      <div className="space-y-2">
+        <label className="text-xs text-editor-text-muted font-medium">Low-Confidence Word Threshold</label>
+        <p className="text-[10px] text-editor-text-muted leading-relaxed">
+          Words with confidence below this value are highlighted with an orange dotted underline.
+          Whisper often gets homophones and proper nouns wrong at low confidence.
+        </p>
+        <div className="flex items-center gap-2">
+          <input
+            type="range"
+            min={0}
+            max={1}
+            step={0.05}
+            value={confidenceThreshold}
+            onChange={(e) => setConfidenceThreshold(Number(e.target.value))}
+            className="flex-1 h-1.5"
+          />
+          <input
+            type="number"
+            min={0}
+            max={1}
+            step={0.05}
+            value={confidenceThreshold}
+            onChange={(e) => setConfidenceThreshold(Math.max(0, Math.min(1, Number(e.target.value) || 0)))}
+            className="w-16 px-2 py-1 bg-editor-bg border border-editor-border rounded-lg text-xs text-editor-text focus:outline-none focus:border-editor-accent"
+          />
+        </div>
+        <div className="flex items-center justify-between text-[10px]">
+          <span className="text-editor-text-muted">Show all</span>
+          <span className="font-medium text-editor-text">{confidenceThreshold.toFixed(2)}</span>
+          <span className="text-editor-text-muted">Strict</span>
+        </div>
+      </div>
+
       {/* Default provider selector */}
       <div className="space-y-2">
         <label className="text-xs text-editor-text-muted font-medium">Default AI Provider</label>

frontend/src/components/TranscriptEditor.tsx

@@ -49,6 +49,10 @@ export default function TranscriptEditor({
   const [searchQuery, setSearchQuery] = useState('');
   const [activeMatchIdx, setActiveMatchIdx] = useState(0);
   const searchInputRef = useRef<HTMLInputElement | null>(null);
+  const updateWordText = useEditorStore((s) => s.updateWordText);
+  const [editingWordIndex, setEditingWordIndex] = useState<number | null>(null);
+  const [editText, setEditText] = useState('');
+  const editInputRef = useRef<HTMLInputElement | null>(null);
 
   const selectedSet = useMemo(() => new Set(selectedWordIndices), [selectedWordIndices]);
   const matchIndices = useMemo(() => {
@@ -224,6 +228,61 @@ export default function TranscriptEditor({
     [setSelectedWordIndices],
   );
 
+  const startEditing = useCallback((index: number) => {
+    const word = words[index];
+    if (!word) return;
+    setEditingWordIndex(index);
+    setEditText(word.word);
+    requestAnimationFrame(() => {
+      editInputRef.current?.focus();
+      editInputRef.current?.select();
+    });
+  }, [words]);
+
+  const commitEdit = useCallback(() => {
+    if (editingWordIndex === null) return;
+    const trimmed = editText.trim();
+    if (trimmed && trimmed !== words[editingWordIndex]?.word) {
+      updateWordText(editingWordIndex, trimmed);
+    }
+    setEditingWordIndex(null);
+    setEditText('');
+  }, [editingWordIndex, editText, words, updateWordText]);
+
+  const cancelEdit = useCallback(() => {
+    setEditingWordIndex(null);
+    setEditText('');
+  }, []);
+
+  const handleWordDoubleClick = useCallback((index: number) => {
+    if (cutMode || muteMode || gainMode || speedMode) return;
+    startEditing(index);
+  }, [cutMode, muteMode, gainMode, speedMode, startEditing]);
+
+  // Focus edit input when it appears
+  useEffect(() => {
+    if (editingWordIndex !== null && editInputRef.current) {
+      editInputRef.current.focus();
+      editInputRef.current.select();
+    }
+  }, [editingWordIndex]);
+
+  // Global key handler for edit mode
+  useEffect(() => {
+    const onKeyDown = (e: KeyboardEvent) => {
+      if (editingWordIndex === null) return;
+      if (e.key === 'Enter') {
+        e.preventDefault();
+        commitEdit();
+      } else if (e.key === 'Escape') {
+        e.preventDefault();
+        cancelEdit();
+      }
+    };
+    window.addEventListener('keydown', onKeyDown);
+    return () => window.removeEventListener('keydown', onKeyDown);
+  }, [editingWordIndex, commitEdit, cancelEdit]);
+
   const cutSelectedWords = useCallback(() => {
     if (selectedWordIndices.length === 0) return;
     const sorted = [...selectedWordIndices].sort((a, b) => a - b);
@@ -319,15 +378,25 @@ export default function TranscriptEditor({
               const isSearchMatch = matchSet.has(globalIndex);
               const isActiveSearchMatch = matchIndices.length > 0 && matchIndices[safeActiveMatchIdx] === globalIndex;
 
+              const isEditing = globalIndex === editingWordIndex;
+
+              // Low-confidence highlighting
+              const CONFIDENCE_THRESHOLD_KEY = 'talkedit:confidenceThreshold';
+              const storedThreshold = typeof window !== 'undefined' ? Number(window.localStorage.getItem(CONFIDENCE_THRESHOLD_KEY)) : 0;
+              const confidenceThreshold = Number.isFinite(storedThreshold) ? storedThreshold : 0.6;
+              const isLowConfidence = word.confidence > 0 && word.confidence < confidenceThreshold && !cutRange && !muteRange && !gainRange && !speedRange;
+              const confidencePct = word.confidence > 0 ? Math.round(word.confidence * 100) : null;
+
               return (
                 <span
                   key={globalIndex}
                   id={`word-${globalIndex}`}
                   data-word-index={globalIndex}
-                  title={`${word.start.toFixed(2)}s — Ctrl+click to seek`}
+                  title={`${word.start.toFixed(2)}s — confidence: ${confidencePct !== null ? confidencePct + '%' : 'N/A'}${isLowConfidence ? ' ⚠️ Low confidence' : ''} — Ctrl+click to seek, double-click to edit`}
                   onMouseDown={(e) => handleWordMouseDown(globalIndex, e)}
                   onMouseEnter={() => handleWordMouseEnter(globalIndex)}
                   onMouseLeave={() => setHoveredWordIndex(null)}
+                  onDoubleClick={() => handleWordDoubleClick(globalIndex)}
                   className={`
                     relative px-[2px] py-[1px] rounded cursor-pointer transition-colors
                     ${cutRange ? 'bg-red-500/20 text-red-100' : ''}
@@ -343,9 +412,21 @@ export default function TranscriptEditor({
                     ${isSelected && !cutRange && !muteRange && !gainRange && !speedRange ? 'bg-editor-word-selected text-white' : ''}
                     ${isActive && !isSelected && !cutRange && !muteRange && !gainRange && !speedRange ? 'bg-editor-accent/20 text-editor-accent' : ''}
                     ${isHovered && !isSelected && !isActive && !cutRange && !muteRange && !gainRange && !speedRange ? 'bg-editor-word-hover' : ''}
+                    ${isLowConfidence ? 'border-b border-dashed border-orange-400/60' : ''}
                   `}
                 >
-                  {word.word}{' '}
+                  {isEditing ? (
+                    <input
+                      ref={editInputRef}
+                      value={editText}
+                      onChange={(e) => setEditText(e.target.value)}
+                      onBlur={commitEdit}
+                      className="w-24 px-1 py-0 text-xs bg-editor-bg border border-editor-accent rounded text-editor-text focus:outline-none"
+                      style={{ minWidth: `${Math.max(word.word.length * 8, 48)}px` }}
+                    />
+                  ) : (
+                    <>{word.word}{' '}</>
+                  )}
                   {(cutRange || muteRange || gainRange || speedRange) && isHovered && (
                     <button
                       onClick={(e) => {
@@ -367,7 +448,7 @@ export default function TranscriptEditor({
         </div>
       );
     },
-    [segments, selectedSet, matchSet, matchIndices, safeActiveMatchIdx, activeWordIndex, hoveredWordIndex, handleWordMouseDown, handleWordMouseEnter, setHoveredWordIndex, getCutRangeForWord, getMuteRangeForWord, getGainRangeForWord, getSpeedRangeForWord, removeCutRange, removeMuteRange, removeGainRange, removeSpeedRange, zoneDragRange, cutMode, muteMode, gainMode, speedMode],
+    [segments, selectedSet, matchSet, matchIndices, safeActiveMatchIdx, activeWordIndex, hoveredWordIndex, handleWordMouseDown, handleWordMouseEnter, setHoveredWordIndex, getCutRangeForWord, getMuteRangeForWord, getGainRangeForWord, getSpeedRangeForWord, removeCutRange, removeMuteRange, removeGainRange, removeSpeedRange, zoneDragRange, cutMode, muteMode, gainMode, speedMode, editingWordIndex, editText, editInputRef, handleWordDoubleClick, commitEdit, setEditText],
   );
 
   return (

frontend/src/lib/tauri-bridge.ts

@@ -11,7 +11,7 @@ import { invoke } from '@tauri-apps/api/core';
 import { open, save } from '@tauri-apps/plugin-dialog';
 import { readTextFile, writeTextFile } from '@tauri-apps/plugin-fs';
 
-const backendPort = import.meta.env.VITE_BACKEND_PORT || '8642';
+const backendPort = import.meta.env.VITE_BACKEND_PORT || '8000';
 const backendUrl = `http://127.0.0.1:${backendPort}`;
 
 const VIDEO_FILTERS = [

frontend/src/store/editorStore.ts

@@ -67,6 +67,7 @@ interface EditorActions {
   setHoveredWordIndex: (index: number | null) => void;
   deleteSelectedWords: () => void;
   deleteWordRange: (startIndex: number, endIndex: number) => void;
+  updateWordText: (index: number, text: string) => void;
   addCutRange: (start: number, end: number, trimGroupId?: string) => void;
   addMuteRange: (start: number, end: number) => void;
   addGainRange: (start: number, end: number, gainDb: number) => void;
@@ -262,6 +263,33 @@ export const useEditorStore = create<EditorState & EditorActions>()(
         get().addCutRange(words[startIndex].start, words[endIndex].end);
       },
 
+      updateWordText: (index, text) => {
+        const { words, segments } = get();
+        if (index < 0 || index >= words.length) return;
+        const newWords = words.map((w, i) =>
+          i === index ? { ...w, word: text } : w
+        );
+        // Also update the corresponding segment's words and text
+        let globalIdx = 0;
+        const newSegments = segments.map((seg) => {
+          const start = globalIdx;
+          globalIdx += seg.words.length;
+          if (index >= start && index < start + seg.words.length) {
+            const localIdx = index - start;
+            const updatedSegWords = seg.words.map((w, i) =>
+              i === localIdx ? { ...w, word: text } : w
+            );
+            return {
+              ...seg,
+              words: updatedSegWords,
+              text: updatedSegWords.map((w) => w.word).join(' '),
+            };
+          }
+          return seg;
+        });
+        set({ words: newWords, segments: newSegments });
+      },
+
       addCutRange: (start, end, trimGroupId) => {
         const { cutRanges } = get();
         const newRange: CutRange = {

frontend/src/types/project.ts

@@ -87,7 +87,7 @@ export interface ExportOptions {
   outputPath: string;
   mode: 'fast' | 'reencode';
   resolution: '720p' | '1080p' | '4k';
-  format: 'mp4' | 'mov' | 'webm';
+  format: 'mp4' | 'mov' | 'webm' | 'wav';
   enhanceAudio: boolean;
   captions: 'none' | 'burn-in' | 'sidecar';
   captionStyle?: CaptionStyle;