Turn "Skip silence" on by default (opt-out, D-052)

Owner's decision: the long-form pipeline's gain on the long real recording
outweighs the spec's strict every-clip rollout rule. A stored "off" still wins.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

docs/Architecture.md

@@ -168,7 +168,7 @@ environment). See `WorkBreakdown.md` for the itemized checklist.)
    to the decoder input of every 30 s window by a wrapper around the model's
    `generate()`, which also strips it from the output again (D-048).
    Live passes (4.1b) carry the same `prompt`.
-6. **Long-form branch (D-049/D-050, "Advanced → Skip silence", off by default).**
+6. **Long-form branch (D-049/D-050, "Advanced → Skip silence", on by default since D-052).**
    With the option on, every load posted for a file run or the Download button carries
    `vad: true`; once the Whisper pipeline is ready the worker loads Silero VAD
    (`onnx-community/silero-vad`, 2.2 MB, Cache API like the weights; offline it is

docs/Decisions.md

@@ -1629,7 +1629,7 @@ tokenizer incl. "Århus, Ærø"); server-live passes keep the ids. Pinned by the
 stubbed-sidecar test "the vocabulary hint reaches the batched pipeline as text".
 
 ## D-049 — Long-form file pipeline (Silero VAD chunks, language pinned once) ships as an Advanced option, off by default
-**Status:** Accepted (2026-10-30) — spec 001 Goal 3
+**Status:** Accepted (2026-10-30) — spec 001 Goal 3; **the "off by default" part is superseded by D-052**
 Browser file transcription cuts audio into fixed 30 s windows with 5 s
 overlap (Transformers.js's chunked pipeline), decoding silence and
 overlaps, and with "Auto-detect" Transformers.js 4.2 silently decodes as
@@ -1801,3 +1801,23 @@ fallback restart behaves; Firefox/Safari WASM.
 **Revisit:** if Transformers.js gains a TDT decoder, move to it and drop
 parakeet.js + the alias plugin; if WebGPU measures well on real hardware,
 consider letting "auto" pick it.
+
+## D-052 — "Skip silence" (long-form pipeline) is on by default (opt-out)
+**Status:** Accepted (2026-10-30) — owner's decision; supersedes D-049's default
+D-049 shipped the long-form pipeline off by default because the spec's
+rollout rule required it to be faster *and* no worse on WER on every
+measurement clip, and two clips missed: the synthetic meeting (+0.0106 WER,
+mostly casual spellings like "gonna") and da_festuge with `da` (a speed tie).
+On the one long real recording it was ~2× faster with about half the WER
+(en_alien, Large v3 Turbo on CPU: 210 s → 115 s, 0.255 → 0.140), and it
+fixes "auto" decoding Danish as English. The owner weighed that above the
+strict rule and made it opt-out.
+**Change:** `LONG_FORM_DEFAULT = true` in `src/lib/longForm.ts`. A stored
+choice still wins, so anyone who turned it off keeps it off. Nothing else
+changes: the VAD model (2.2 MB, D-050) is still fetched only as part of a
+user-started transcription or the Download button, and when it isn't
+available (offline, never downloaded) the file still goes through the
+chunked path. Live mode is unaffected.
+**Caveat:** the WER numbers behind this use reference transcripts generated
+with Whisper (owner, 2026-10-30), so they favour Whisper-like output; re-check
+once correct transcripts are available.

src/lib/longForm.ts

@@ -8,11 +8,12 @@ import type { StorageLike } from "./engine";
 export const LONG_FORM_KEY = "whisper-long-form";
 
 /**
- * Off by default: the spec's rollout rule makes the long-form pipeline the
- * default only if it is faster and not less accurate on every measurement
- * clip — see D-049 for the numbers that decided this.
+ * On by default (opt-out, D-052): the owner chose it after D-049's
+ * measurements — ~2× faster and about half the WER on the 15-min clip —
+ * outweighed the spec's strict every-clip rule, which kept it opt-in on a
+ * +0.0106 WER result on the synthetic meeting. A stored "off" still wins.
  */
-export const LONG_FORM_DEFAULT = false;
+export const LONG_FORM_DEFAULT = true;
 
 function defaultStorage(): StorageLike | null {
   try {

tests/longform-setting.test.ts

@@ -34,6 +34,12 @@ describe("long-form setting persistence (D-049)", () => {
     expect(() => saveLongFormChoice(true, throwing)).not.toThrow();
   });
 
+  test("is on unless the user turned it off (opt-out, D-052)", () => {
+    expect(LONG_FORM_DEFAULT).toBe(true);
+    expect(loadLongFormChoice(memoryStorage())).toBe(true);
+    expect(loadLongFormChoice(memoryStorage({ [LONG_FORM_KEY]: "off" }))).toBe(false);
+  });
+
   test("round-trips an explicit choice", () => {
     const storage = memoryStorage();
     saveLongFormChoice(!LONG_FORM_DEFAULT, storage);