diff --git a/CURRENT.md b/CURRENT.md
index bb29ed1..c2e09de 100644
--- a/CURRENT.md
+++ b/CURRENT.md
@@ -40,6 +40,12 @@ mobile draft at 390px.
   - [x] C11 preview caching: Discord-hosted image/GIF previews (WebP 320px, hash-addressed, 5 MiB
     cap, one attempt) run after every walk in both modes. Live run `...-20260916090722`:
     `previewed: 2`; thumbs served over `/api/media/<scrape>/<item>/thumb` (200).
+  - [x] C69 Streaming silence fix. Word mode with streaming on played nothing but the hum: the
+    daemon started ffplay as soon as the streamed file passed 44 bytes (the bare WAV header), so
+    the player hit EOF and exited before the first word landed. The wrapper now touches
+    `<file>.ready` after the first word, and `playWhenReady` waits for it (size threshold and
+    timeout fallbacks for other providers). Verified by capturing `veskslop_tts.monitor`: peak
+    -11.2 dBFS with 7 word gaps, versus -48 dBFS (hum only) before.
   - [x] C68 Half-Life voices and a tidier Radio kit. `vox` and `hev` are now real voices in the
     list (Keith Bell with the VOX/HEV profile; VOX implies its prosody), so the preset dropdown and
     the VOX/clean toggles left the Radio kit; the clean option appears only for the VOX voice, and
diff --git a/docs/voice.md b/docs/voice.md
index 8498633..ab569cf 100644
--- a/docs/voice.md
+++ b/docs/voice.md
@@ -170,7 +170,9 @@ from the TTS panel:
   s16le, so only the character changes. Echo gained `tts.echoDelayMs` (10-400) and
   `tts.echoLevel` (0-0.95) alongside the amount. **Stream word mode** (`tts.streamWords`, on by
   default) makes the wrapper write the WAV while it synthesizes, so playback starts on the first
-  word: measured first sound 0.5 s instead of 2.4 s on a ten-word line. Streaming skips the
+  word: measured first sound 0.5 s instead of 2.4 s on a ten-word line. The wrapper touches
+  `<file>.ready` after the first word and the daemon waits for that before starting playback, so a
+  player never opens the bare WAV header and exits before the words arrive. Streaming skips the
   loudness measurement (`VESKSLOP_TTS_STREAM_GAIN_DB` compensates if needed) and cannot use VOX
   prosody, which needs the whole sentence.
   The Half-Life voices are first-class entries in the voice list: **VOX (Half-Life)** and **HEV
diff --git a/microslopsam/setup/veskslop-tts.sh b/microslopsam/setup/veskslop-tts.sh
index 914e6f3..8dafc79 100755
--- a/microslopsam/setup/veskslop-tts.sh
+++ b/microslopsam/setup/veskslop-tts.sh
@@ -602,6 +602,9 @@ if [ "$STREAM_WORDS" = "1" ]; then
         fi
         cat "$TMP/sw$INDEX.raw" >> "$OUT" 2>/dev/null || { echo "stream write failed" >&2; exit 1; }
 
+        # The player waits for this marker so it does not start on the bare header.
+        if [ "$INDEX" = "1" ]; then : > "$OUT.ready"; fi
+
         if [ "$GAP" -gt 0 ] && [ "$INDEX" -lt "$WORD_COUNT" ]; then
             if "$MICROSLOP_FFMPEG" -hide_banner -loglevel error -y -f lavfi -i "anullsrc=r=48000:cl=mono" \
                     -t "$(awk -v g="$GAP" 'BEGIN{printf "%.3f", g/1000}')" \
diff --git a/poc/src/audio-service.ts b/poc/src/audio-service.ts
index e125282..031c899 100644
--- a/poc/src/audio-service.ts
+++ b/poc/src/audio-service.ts
@@ -6,7 +6,7 @@
  * them in the transcript, and runs speech to text when a provider is set.
  */
 
-import { statSync } from "node:fs";
+import { statSync, unlinkSync } from "node:fs";
 import { AudioManager } from "./audio.js";
 import type { CaptureSegment } from "./capture.js";
 import { CapturePipeline } from "./capture.js";
@@ -147,17 +147,35 @@ export class AudioService {
         return { queued: true, id, position: this.queue.length };
     }
 
-    /** Wait for the wrapper's streaming file to appear, then play it while it grows. */
+    /**
+     * Wait until the wrapper's streaming file has real audio in it, then play it while it
+     * grows. The wrapper touches `<file>.ready` after the first word, which is the reliable
+     * signal; a size threshold and a timeout cover providers that do not mark readiness.
+     */
     private async playWhenReady(file: string): Promise<void> {
         for (let attempt = 0; attempt < 400; attempt++) {
             try {
-                if (statSync(file).size > 44) break;
+                if (statSync(`${file}.ready`).size >= 0) break;
+            } catch {
+                // no marker yet
+            }
+            try {
+                if (statSync(file).size > 100_000) break;
             } catch {
                 // not created yet
             }
             await new Promise(resolve => setTimeout(resolve, 25));
         }
-        await this.manager.play(file);
+
+        try {
+            await this.manager.play(file);
+        } finally {
+            try {
+                unlinkSync(`${file}.ready`);
+            } catch {
+                // best effort
+            }
+        }
     }
 
     stop(): void {