docs: Audio workspace tab (Phase 1) — v1.3 changelog + README; drop orphaned transport spacer

README.md

@@ -33,6 +33,14 @@ All clips are exactly 8 seconds — the standard length for foley sound datasets
 - **Hardware encoding** — GPU-accelerated export via NVENC, VAAPI, QSV, AMF, or VideoToolbox
 - **Subject tracking** — auto-adjust crop center using YOLOv8 detection (optional)
 
+### Audio extraction & editing
+
+- **Extract audio area** — save an exact-length audio slice from the playhead (length steps by 1s, no upper cap)
+- **Output format** — WAV / MP3 / FLAC / M4A / OGG / OPUS, remembered between sessions
+- **Non-destructive edits** — fade in/out, EBU R128 normalize, and gain, applied via ffmpeg on extract (and audition)
+- **Waveform preview** — a read-only strip of the current area (↻ to refresh; decode length-capped so long areas don't stall)
+- **Audition** — Play/Stop the current area with the edits applied
+
 ### Audio scanning
 
 - **Embedding models** — WAV2VEC2 (base/large), HuBERT (base/large/xlarge), BEATs
@@ -43,6 +51,7 @@ All clips are exactly 8 seconds — the standard length for foley sound datasets
 - **Hard negatives** — mark false positives to refine training
 - **Model versioning** — timestamped backups with rollback support
 - **Scan export** — batch export from scan results with spread and minimum duration filtering
+- **Location** — scan controls live in the **Audio** tab's *Scan / Classify* pane (formerly the *Scan* tab)
 
 ### Scan results panel