diff --git a/web/zebra-spaces.html b/web/zebra-spaces.html
index d4a1e37..74bdff3 100644
--- a/web/zebra-spaces.html
+++ b/web/zebra-spaces.html
@@ -183,6 +183,31 @@
     position: relative;
   }
   .controls { min-width: 0; }
+  /* Transcript log — lives in the timeline column under the
+   * spotlight. Off by default (the section is .hidden); toggled
+   * visible via the controls-column transcribe button. Auto-scrolls
+   * to bottom unless the user scrolls up to read older lines. */
+  #sec-transcript {
+    margin-top: 0.6rem; border: 1px solid #ddd; background: #fafafa;
+    padding: 0.4rem 0.6rem; font-size: 0.85rem; min-width: 0;
+  }
+  html.theme-dark #sec-transcript { background: #0d0d0d; border-color: #333; }
+  #transcript-log {
+    max-height: 14rem; overflow-y: auto; line-height: 1.45;
+    font-family: ui-monospace, monospace; white-space: pre-wrap;
+    word-break: break-word;
+  }
+  .transcript-line { padding: 0.1rem 0; border-bottom: 1px dotted #e5e5e5; }
+  html.theme-dark .transcript-line { border-color: #222; }
+  .transcript-line:last-child { border-bottom: none; }
+  .transcript-line .ts   { color: #888; font-size: 0.75rem; }
+  .transcript-line .name { color: #050; font-weight: bold; }
+  html.theme-dark .transcript-line .name { color: #6cc06c; }
+  .transcript-line .txt  { color: inherit; }
+  /* Toggle button — same shape as other small buttons, with an
+   * 'on' state so the user can see at a glance. */
+  #btn-transcribe.on { background: #050; color: #fff; border-color: #050; }
+  html.theme-dark #btn-transcribe.on { background: #6cc06c; color: #000; border-color: #6cc06c; }
   /* games live in the LEFT column under cameras + screens-thumbs as
    * proper tile-thumbs. Click one to spotlight it in the middle (an
    * iframe loads at full size); the click also broadcasts the spotlight
@@ -651,6 +676,12 @@ try {
   <section id="sec-spotlight" class="hidden">
     <div id="spotlight"></div>
   </section>
+  <!-- transcript log: per-speaker live captions, off by default,
+       toggled from the controls column. Hidden until the user opts in
+       — no model download or audio capture until they ask for it. -->
+  <section id="sec-transcript" class="hidden">
+    <div id="transcript-log" aria-live="polite"></div>
+  </section>
 </main>
 
 <aside class="controls">
@@ -715,6 +746,10 @@ try {
   <div class="row" id="row-music-mode">
     <label class="note"><input type="checkbox" id="music-mode"> music mode — raw mic, no echo/noise cancellation (for playing audio through it)</label>
   </div>
+  <div class="row">
+    <button id="btn-transcribe" class="small">transcribe (off)</button>
+    <span class="note">live captions per speaker, on-device whisper-tiny.en (~40 MB once, cached). nothing leaves your browser.</span>
+  </div>
   <p class="note">join as listener; host promotes to mic. end-to-end encrypted.</p>
 </section>
 
@@ -1581,6 +1616,7 @@ function renderIdentity(){
 }
 
 $('btn-vault').addEventListener('click', () => $('vault-panel').classList.toggle('hidden'));
+$('btn-transcribe').addEventListener('click', () => toggleTranscribe().catch(e => logLine('err','transcribe toggle: '+e.message)));
 
 /* log out — destructive: wipes Ed25519 + handle from localStorage and
  * generates a fresh identity. The booted/blocked window keys off the
@@ -1950,6 +1986,196 @@ function loadJitterWorklet(ctx){
     logLine('err', 'jitter-buffer worklet load: '+e.message+' — listener audio direct');
   });
 }
+
+/* ==================================================================
+ * Client-side speech-to-text — whisper-tiny.en via transformers.js
+ *
+ * Off by default. Toggleable in the UI. When on, every audible speaker
+ * gets a parallel AudioWorklet capture node that decimates 48kHz mic
+ * audio down to 16kHz mono, batches 5-second chunks, and ships them
+ * to a Whisper pipeline running in the main thread. Recognized text
+ * is appended to a transcript log under the spotlight as
+ * "HH:MM:SS  name: text".
+ *
+ * Privacy: model runs entirely in the listener's browser — no audio
+ * leaves the device for transcription. First ON downloads ~40MB
+ * model (cached afterward). Subsequent ON is instant.
+ *
+ * Fox 2026-06-05: "implement this and please this is perfect, make it
+ * an off be default toggle that is part of the client. individual
+ * speakers should use their names and show up like a log under the
+ * video area in middle." */
+const WHISPER_CAPTURE_WORKLET_CODE = `
+class WhisperCaptureProcessor extends AudioWorkletProcessor {
+  constructor(){
+    super();
+    this.enabled = false;
+    /* 5 seconds at 16kHz target — sweet spot for whisper-tiny.en
+     * accuracy without too much latency. */
+    this.chunkTargetSamples = 16000 * 5;
+    this.outRate = 16000;
+    this.decimationFactor = Math.round(sampleRate / this.outRate);
+    if (this.decimationFactor < 1) this.decimationFactor = 1;
+    this.decimateCounter = 0;
+    this.buf = new Float32Array(this.chunkTargetSamples + 4096);
+    this.bufPos = 0;
+    this.port.onmessage = (e) => {
+      if (!e.data) return;
+      if (e.data.cmd === 'start') this.enabled = true;
+      else if (e.data.cmd === 'stop'){
+        this.enabled = false;
+        this.bufPos = 0;
+        this.decimateCounter = 0;
+      }
+    };
+  }
+  process(inputs){
+    if (!this.enabled) return true;
+    const ch = inputs[0] && inputs[0][0];
+    if (!ch || ch.length === 0) return true;
+    for (let i = this.decimateCounter; i < ch.length; i += this.decimationFactor){
+      if (this.bufPos < this.buf.length) this.buf[this.bufPos++] = ch[i];
+    }
+    this.decimateCounter = (this.decimateCounter + ch.length) % this.decimationFactor;
+    if (this.bufPos >= this.chunkTargetSamples){
+      const chunk = this.buf.slice(0, this.bufPos);
+      this.port.postMessage({ chunk }, [chunk.buffer]);
+      this.bufPos = 0;
+    }
+    return true;
+  }
+}
+registerProcessor('whisper-capture', WhisperCaptureProcessor);
+`;
+
+let whisperCaptureWorkletReady = false;
+let whisperCaptureWorkletLoading = false;
+function loadWhisperCaptureWorklet(ctx){
+  if (whisperCaptureWorkletReady || whisperCaptureWorkletLoading) return Promise.resolve(whisperCaptureWorkletReady);
+  whisperCaptureWorkletLoading = true;
+  const blob = new Blob([WHISPER_CAPTURE_WORKLET_CODE], { type: 'application/javascript' });
+  const url = URL.createObjectURL(blob);
+  return ctx.audioWorklet.addModule(url).then(() => {
+    URL.revokeObjectURL(url);
+    whisperCaptureWorkletReady = true;
+    whisperCaptureWorkletLoading = false;
+    return true;
+  }).catch(e => {
+    URL.revokeObjectURL(url);
+    whisperCaptureWorkletLoading = false;
+    logLine('err', 'whisper-capture worklet load: '+e.message);
+    return false;
+  });
+}
+
+let whisperPipeline = null;
+let whisperLoading = false;
+async function ensureWhisperPipeline(){
+  if (whisperPipeline) return whisperPipeline;
+  if (whisperLoading){
+    while (whisperLoading && !whisperPipeline) await new Promise(r => setTimeout(r, 200));
+    return whisperPipeline;
+  }
+  whisperLoading = true;
+  logLine('', 'whisper: loading model (~40MB, first time only)…');
+  try {
+    /* transformers.js from jsDelivr — runs ONNX Runtime + whisper-tiny.en
+     * fully in-browser. Model is downloaded once and cached by the
+     * service worker / browser cache; subsequent toggles are instant. */
+    const tx = await import('https://cdn.jsdelivr.net/npm/@huggingface/transformers@3.0.0');
+    whisperPipeline = await tx.pipeline('automatic-speech-recognition', 'Xenova/whisper-tiny.en');
+    logLine('', 'whisper: model ready');
+  } catch (e){
+    logLine('err', 'whisper load failed: '+e.message);
+  } finally {
+    whisperLoading = false;
+  }
+  return whisperPipeline;
+}
+
+let transcribeEnabled = false;
+function appendTranscriptLine(uuid, text){
+  const log = document.getElementById('transcript-log');
+  if (!log) return;
+  const member = members.get(uuid);
+  const name = (member && member.handle) ? member.handle : uuid.slice(0,4);
+  const now = new Date();
+  const ts = String(now.getHours()).padStart(2,'0')+':'+
+             String(now.getMinutes()).padStart(2,'0')+':'+
+             String(now.getSeconds()).padStart(2,'0');
+  const line = document.createElement('div');
+  line.className = 'transcript-line';
+  const tsEl = document.createElement('span'); tsEl.className = 'ts'; tsEl.textContent = ts;
+  const nameEl = document.createElement('span'); nameEl.className = 'name'; nameEl.textContent = name + ':';
+  const txtEl = document.createElement('span'); txtEl.className = 'txt'; txtEl.textContent = text;
+  line.appendChild(tsEl); line.append(' ');
+  line.appendChild(nameEl); line.append(' ');
+  line.appendChild(txtEl);
+  log.appendChild(line);
+  /* auto-scroll only if user is already at the bottom (let them scroll
+   * up to read older transcript without being yanked back). */
+  const nearBottom = (log.scrollTop + log.clientHeight) >= (log.scrollHeight - 40);
+  if (nearBottom) log.scrollTop = log.scrollHeight;
+}
+
+async function startCaptureForUuid(uuid){
+  const node = listenerAudioNodes.get(uuid);
+  if (!node || !audioCtx || node.capture) return;
+  const ok = await loadWhisperCaptureWorklet(audioCtx);
+  if (!ok) return;
+  try {
+    const capture = new AudioWorkletNode(audioCtx, 'whisper-capture');
+    capture.port.onmessage = async (e) => {
+      if (!e.data || !e.data.chunk) return;
+      const pipeline = await ensureWhisperPipeline();
+      if (!pipeline) return;
+      try {
+        const result = await pipeline(e.data.chunk, { sampling_rate: 16000 });
+        const txt = (result && result.text ? result.text : '').trim();
+        /* skip empty / placeholder transcriptions ("." " " "[BLANK_AUDIO]" etc.) */
+        if (txt.length >= 2 && !/^\[/.test(txt)) appendTranscriptLine(uuid, txt);
+      } catch (err){
+        logLine('err', 'whisper transcribe '+uuid.slice(0,4)+': '+err.message);
+      }
+    };
+    /* tap the source — capture runs in parallel with the worklet
+     * chain, doesn't need to connect to destination (we just want
+     * the chunks via port messages). */
+    node.src.connect(capture);
+    capture.port.postMessage({ cmd: 'start' });
+    node.capture = capture;
+    logLine('', 'whisper: capture started for '+uuid.slice(0,4));
+  } catch (e){
+    logLine('err', 'whisper capture install '+uuid.slice(0,4)+': '+e.message);
+  }
+}
+
+function stopCaptureForUuid(uuid){
+  const node = listenerAudioNodes.get(uuid);
+  if (!node || !node.capture) return;
+  try { node.capture.port.postMessage({ cmd: 'stop' }); } catch(_){}
+  try { node.capture.disconnect(); } catch(_){}
+  try { node.src.disconnect(node.capture); } catch(_){}
+  node.capture = null;
+}
+
+async function toggleTranscribe(){
+  transcribeEnabled = !transcribeEnabled;
+  const btn = document.getElementById('btn-transcribe');
+  const sec = document.getElementById('sec-transcript');
+  if (btn) btn.classList.toggle('on', transcribeEnabled);
+  if (btn) btn.textContent = transcribeEnabled ? 'transcribe (on)' : 'transcribe (off)';
+  if (sec) sec.classList.toggle('hidden', !transcribeEnabled);
+  if (transcribeEnabled){
+    /* pre-warm the pipeline so the first chunk doesn't wait. */
+    ensureWhisperPipeline().catch(()=>{});
+    for (const [uuid] of listenerAudioNodes){
+      startCaptureForUuid(uuid).catch(e => logLine('err', 'transcribe start '+uuid+': '+e.message));
+    }
+  } else {
+    for (const [uuid] of listenerAudioNodes) stopCaptureForUuid(uuid);
+  }
+}
 function installJitterBuffer(uuid, node){
   if (!node || node.jbuf || !workletReady) return;
   const target = node.targetSeconds || RECV_PLAYOUT_DELAY_SEC;
@@ -2426,6 +2652,12 @@ function attachAudioStreamViaWorklet(uuid, stream, targetSeconds){
    * existing stream is swapped through the buffer (see loadJitterWorklet) */
   loadJitterWorklet(audioCtx);
   if (workletReady) installJitterBuffer(uuid, node);
+  /* If transcribe is on, install a parallel capture node for this
+   * new speaker so their voice immediately starts flowing into the
+   * transcript log. */
+  if (transcribeEnabled){
+    startCaptureForUuid(uuid).catch(e => logLine('err', 'transcribe '+uuid.slice(0,4)+': '+e.message));
+  }
   logLine('', 'audio via AudioContext '+uuid.slice(0,4)+' target='+targetSeconds+'s ctxState='+audioCtx.state);
   return true;
 }
@@ -2467,6 +2699,7 @@ function detachListenerStream(uuid){
   if (!node) return;
   try { node.src.disconnect(); } catch(_){}
   try { if (node.jbuf) node.jbuf.disconnect(); } catch(_){}
+  try { if (node.capture) node.capture.disconnect(); } catch(_){}
   try { node.gain.disconnect(); } catch(_){}
   listenerAudioNodes.delete(uuid);
 }
@@ -6848,8 +7081,8 @@ logLine('', 'ready — pick a handle, type a rendezvous code, enter the space');
 
 <footer style="margin:2.2rem auto 0;font-size:0.65rem;color:#999;line-height:1.7;word-break:break-all;font-family:monospace">
   <span id="pi-seal" style="color:#777;cursor:default;user-select:none" title="">page integrity</span> &nbsp;·&nbsp; built <span class="stamp-date">2026-06-05</span><br>
-  md5 <span class="stamp-md5">9728607e6d8187fed0bfa947f93cccaa</span><br>
-  sha256 <span class="stamp-sha">26680e413131920d5c0dc981b20460b1c087d08193f1f5d89925d22e3013ec11</span><br>
+  md5 <span class="stamp-md5">4591b92519ec28db69de15bb20eb2300</span><br>
+  sha256 <span class="stamp-sha">0b792e6dd0d9bc69f888c239468c57506bc1d0feb9ded8b78bb650b886501c9b</span><br>
   <span style="color:#bbb">hashes are of this page with these two fields zeroed — to verify, blank them and re-hash</span><br>
   <span style="color:#bbb">one self-contained file — <strong>save a copy</strong> and verify against these hashes; point at your own servers with ?signal= and ?turncred=, or <a href="host-your-own.html" style="color:#999">host your own community</a></span>
 </footer>