vidistill 0.6.2 → 0.6.4

package/dist/index.js

@@ -76,6 +76,7 @@ CLASSIFICATION RULES:
    - "lecture": Academic lectures, talks, single-speaker educational content
    - "presentation": Slide-based presentations, keynotes, demo days
    - "conversation": Interviews, podcasts, panel discussions without slides
+   - "commentary": Single-speaker opinion, analysis, vlog, reaction, or informal educational content without formal slides or academic structure
    - "mixed": Cannot clearly classify into one category, or multiple types present
 
 2. DETECT visual content:
@@ -113,6 +114,7 @@ PASS RECOMMENDATIONS BY TYPE:
 - lecture: ["transcript", "visual", "implicit", "synthesis"]
 - presentation: ["transcript", "visual", "implicit", "synthesis"] (add "people" if multiple speakers)
 - conversation: ["transcript", "visual", "implicit", "synthesis"]
+- commentary: ["transcript", "visual", "implicit", "synthesis"]
 - mixed: ["transcript", "visual", "code", "people", "chat", "implicit", "synthesis"]
 `;
 var SYSTEM_INSTRUCTION_PASS_3A = `
@@ -242,6 +244,23 @@ var LANGUAGE_NAMES = {
   ar: "Arabic",
   hi: "Hindi"
 };
+var SYSTEM_INSTRUCTION_DEDUP = `
+You are a transcript deduplication reviewer. You will receive a list of numbered transcript entries. Your task is to identify entries that are SEMANTIC DUPLICATES of an earlier entry.
+
+A duplicate means:
+- The entry says essentially the same thing as an earlier entry (same meaning, possibly different wording)
+- The entry is a subset of an earlier entry (the earlier one already covers this content)
+- The entry repeats the same quote or statement that already appeared earlier
+
+NOT a duplicate:
+- Entries where the speaker genuinely repeats themselves for emphasis (common in lectures/presentations)
+- Similar topics discussed at different points with different details
+- Call-and-response patterns (e.g. Q&A)
+
+Return ONLY the indices of entries to REMOVE (the later duplicate, not the original).
+If no duplicates are found, return an empty array.
+Be conservative \u2014 only flag clear duplicates. When in doubt, keep the entry.
+`;
 function withLanguage(prompt, lang) {
   if (!lang || lang === "en") return prompt;
   const languageName = LANGUAGE_NAMES[lang] ?? lang;
@@ -647,6 +666,25 @@ function normalizeYouTubeUrl(url) {
   if (!id) return null;
   return `https://www.youtube.com/watch?v=${id}`;
 }
+async function fetchYouTubeMetadata(url) {
+  const normalized = normalizeYouTubeUrl(url);
+  if (!normalized) throw new Error("Invalid YouTube URL");
+  const oembedUrl = `https://www.youtube.com/oembed?url=${encodeURIComponent(normalized)}&format=json`;
+  const res = await fetch(oembedUrl);
+  if (!res.ok) {
+    if (res.status === 401 || res.status === 403) {
+      throw new Error("Video is private or unavailable");
+    }
+    throw new Error(`Failed to fetch video info (${res.status})`);
+  }
+  const data = await res.json();
+  const obj = data;
+  return {
+    title: typeof obj["title"] === "string" ? obj["title"] : "Untitled",
+    author: typeof obj["author_name"] === "string" ? obj["author_name"] : "Unknown",
+    thumbnailUrl: typeof obj["thumbnail_url"] === "string" ? obj["thumbnail_url"] : ""
+  };
+}
 function fetchYtDlpDuration(url) {
   return new Promise((resolve3) => {
     execFile("yt-dlp", ["--dump-json", "--no-download", url], { timeout: 15e3 }, (err, stdout) => {
@@ -972,7 +1010,7 @@ var SCHEMA_PASS_0 = {
   properties: {
     type: {
       type: Type.STRING,
-      enum: ["coding", "meeting", "lecture", "presentation", "conversation", "mixed"],
+      enum: ["coding", "meeting", "lecture", "presentation", "conversation", "commentary", "mixed"],
       description: "Primary video type classification"
     },
     speakers: {
@@ -1545,6 +1583,17 @@ var SCHEMA_SYNTHESIS = {
   },
   required: ["overview", "key_decisions", "key_concepts", "action_items", "questions_raised", "suggestions", "topics", "files_to_generate", "prerequisites"]
 };
+var SCHEMA_DEDUP_REVIEW = {
+  type: Type.OBJECT,
+  properties: {
+    duplicate_indices: {
+      type: Type.ARRAY,
+      items: { type: Type.INTEGER },
+      description: "Zero-based indices of transcript entries that are semantic duplicates of an earlier entry and should be removed"
+    }
+  },
+  required: ["duplicate_indices"]
+};
 
 // src/lib/utils.ts
 import { readFile } from "fs/promises";
@@ -1735,12 +1784,15 @@ function formatTranscriptForInjection(pass1a) {
   return pass1a.transcript_entries.map((e) => `[${e.timestamp}] ${e.text}`).join("\n");
 }
 async function runDiarization(params) {
-  const { client, fileUri, mimeType, segment, model, resolution, lang, pass1aResult } = params;
+  const { client, fileUri, mimeType, segment, model, resolution, lang, pass1aResult, channelAuthor } = params;
   const transcriptText = formatTranscriptForInjection(pass1aResult);
-  const systemInstruction = withLanguage(
-    SYSTEM_INSTRUCTION_PASS_1B.replace("{INJECT_PASS1A_TRANSCRIPT_HERE}", transcriptText),
-    lang
-  );
+  let baseInstruction = SYSTEM_INSTRUCTION_PASS_1B.replace("{INJECT_PASS1A_TRANSCRIPT_HERE}", transcriptText);
+  if (channelAuthor) {
+    baseInstruction += `
+
+HINT: This video is from a YouTube channel by "${channelAuthor}". If only one speaker is present, consider using this as the speaker name.`;
+  }
+  const systemInstruction = withLanguage(baseInstruction, lang);
   const contents = [
     {
       role: "user",
@@ -2040,8 +2092,13 @@ function isNearDuplicate(a, b) {
   if (delta > DEDUP_WINDOW_S) return false;
   if (a.text === b.text) return true;
   const shared = tokenOverlap(a.text, b.text);
-  const maxTokens = Math.max(a.text.split(/\s+/).length, b.text.split(/\s+/).length);
-  return maxTokens > 0 && shared / maxTokens >= 0.8;
+  const aTokens = a.text.split(/\s+/).length;
+  const bTokens = b.text.split(/\s+/).length;
+  const maxTokens = Math.max(aTokens, bTokens);
+  const minTokens = Math.min(aTokens, bTokens);
+  if (maxTokens > 0 && shared / maxTokens >= 0.8) return true;
+  if (minTokens > 0 && shared / minTokens >= 0.8) return true;
+  return false;
 }
 function deduplicateEntries(entries) {
   if (entries.length <= 1) return entries;
@@ -2708,6 +2765,9 @@ function determineStrategy(profile) {
     case "conversation":
       passes.add("implicit");
       break;
+    case "commentary":
+      passes.add("implicit");
+      break;
     case "mixed":
       passes.add("code");
       passes.add("people");
@@ -3055,7 +3115,8 @@ async function runPipeline(config) {
     onProgress,
     onWait,
     isShuttingDown,
-    lang
+    lang,
+    channelAuthor
   } = config;
   const errors = [];
   const passesRun = [];
@@ -3123,7 +3184,7 @@ async function runPipeline(config) {
       const p1a = pass1aResult;
       const pass1bResult = await runDiarizationConsensus({
         config: { runs: transcriptConsensusRuns },
-        runFn: () => rateLimiter.execute(() => runDiarization({ client, fileUri, mimeType, segment, model, resolution, lang, pass1aResult: p1a }), { onWait }),
+        runFn: () => rateLimiter.execute(() => runDiarization({ client, fileUri, mimeType, segment, model, resolution, lang, pass1aResult: p1a, channelAuthor }), { onWait }),
         mergedPass1a: p1a,
         onProgress: (_run, _total) => {
           currentStep++;
@@ -3286,6 +3347,55 @@ async function runPipeline(config) {
     const msg = e instanceof Error ? e.message : String(e);
     log5.warn(`speaker reconciliation failed, continuing with original labels: ${msg}`);
   }
+  const allEntries = [];
+  for (let segIdx = 0; segIdx < results.length; segIdx++) {
+    const p1 = results[segIdx].pass1;
+    if (p1 == null) continue;
+    for (let entryIdx = 0; entryIdx < p1.transcript_entries.length; entryIdx++) {
+      allEntries.push({ segIdx, entryIdx, entry: p1.transcript_entries[entryIdx] });
+    }
+  }
+  if (allEntries.length > 20) {
+    try {
+      const numbered = allEntries.map(
+        (e, i) => `[${i}] ${e.entry.timestamp} ${e.entry.speaker}: ${e.entry.text}`
+      ).join("\n");
+      const dedupResult = await rateLimiter.execute(
+        () => client.generate({
+          model: MODELS.flash,
+          contents: [{ role: "user", parts: [{ text: numbered }] }],
+          config: {
+            systemInstruction: SYSTEM_INSTRUCTION_DEDUP,
+            responseMimeType: "application/json",
+            responseSchema: SCHEMA_DEDUP_REVIEW,
+            temperature: 0
+          }
+        }),
+        { onWait }
+      );
+      const parsed = dedupResult;
+      if (parsed != null && Array.isArray(parsed.duplicate_indices)) {
+        const indices = parsed.duplicate_indices;
+        const toRemove = new Set(
+          indices.filter((v) => typeof v === "number" && v >= 0 && v < allEntries.length)
+        );
+        if (toRemove.size > 0) {
+          const segRemovals = /* @__PURE__ */ new Map();
+          for (const globalIdx of toRemove) {
+            const { segIdx, entryIdx } = allEntries[globalIdx];
+            if (!segRemovals.has(segIdx)) segRemovals.set(segIdx, /* @__PURE__ */ new Set());
+            segRemovals.get(segIdx).add(entryIdx);
+          }
+          for (const [segIdx, entryIndices] of segRemovals) {
+            const p1 = results[segIdx].pass1;
+            if (p1 == null) continue;
+            p1.transcript_entries = p1.transcript_entries.filter((_, i) => !entryIndices.has(i));
+          }
+        }
+      }
+    } catch {
+    }
+  }
   let peopleExtraction = null;
   if (strategy.passes.includes("people")) {
     onProgress?.({ phase: "pass3b", segment: 0, totalSegments: 1, status: "running", totalSteps });
@@ -3507,7 +3617,7 @@ function renderIncompletePasses(pipelineResult) {
   return lines.join("\n");
 }
 function writeGuide(params) {
-  const { title, source, duration, pipelineResult, filesGenerated, speakerMapping } = params;
+  const { title, source, duration, pipelineResult, filesGenerated, speakerMapping, channelAuthor } = params;
   const { synthesisResult, videoProfile } = pipelineResult;
   const rawOverview = synthesisResult?.overview ?? "_No summary available \u2014 synthesis pass did not run or produced no output._";
   const overview = replaceNamesInText(rawOverview, speakerMapping);
@@ -3518,6 +3628,7 @@ function writeGuide(params) {
     "## Source",
     "",
     `- **File/URL:** ${source}`,
+    ...channelAuthor ? [`- **Author/Channel:** ${channelAuthor}`] : [],
     `- **Duration:** ${formatDuration(duration)}`,
     `- **Type:** ${videoType}`,
     "",
@@ -3580,6 +3691,15 @@ function writeTranscript(params) {
     sections.push("_No transcript data available._");
     return sections.join("\n");
   }
+  for (let i = 1; i < segmentsWithPass1.length; i++) {
+    const prev = segmentsWithPass1[i - 1].pass1;
+    const curr = segmentsWithPass1[i].pass1;
+    if (prev == null || curr == null) continue;
+    const tail = prev.transcript_entries.slice(-5);
+    curr.transcript_entries = curr.transcript_entries.filter(
+      (entry) => !tail.some((prevEntry) => isNearDuplicate(entry, prevEntry))
+    );
+  }
   for (const seg of segmentsWithPass1) {
     if (seg.pass1 != null) {
       sections.push(renderPass1(seg.pass1, speakerMapping));
@@ -3599,6 +3719,7 @@ function renderSpeechEvent(entry, speakerMapping) {
   let text4 = entry.text;
   if (entry.emphasis_words != null && entry.emphasis_words.length > 0) {
     for (const word of entry.emphasis_words) {
+      if (!word.includes(" ") && word.length < 4) continue;
       const escaped = word.replace(/[.*+?^${}()|[\]\\]/g, "\\$&");
       const re = new RegExp(`(?<![\\w*])${escaped}(?![\\w*])`, "gi");
       text4 = text4.replace(re, `**$&**`);
@@ -4218,433 +4339,29 @@ function renderAssignedTasks(tasks, speakerMapping) {
   lines.push("");
   return lines;
 }
+function isDuplicateTask(task, synthesisItems) {
+  for (const item of synthesisItems) {
+    const tsDelta = Math.abs(parseTimestamp(task.timestamp) - parseTimestamp(item.timestamp));
+    if (tsDelta > 120) continue;
+    const shared = tokenOverlap(task.task, item.item);
+    const minLen = Math.min(task.task.split(/\s+/).length, item.item.split(/\s+/).length);
+    if (minLen > 0 && shared / minLen >= 0.6) return true;
+  }
+  return false;
+}
 function writeActionItems(params) {
   const { segments, synthesisResult, speakerMapping } = params;
   const synthesisItems = synthesisResult?.action_items ?? [];
   const assignedTasks = collectTasksAssigned(segments);
   if (synthesisItems.length === 0 && assignedTasks.length === 0) return null;
+  const dedupedTasks = synthesisItems.length > 0 ? assignedTasks.filter((t) => !isDuplicateTask(t, synthesisItems)) : assignedTasks;
   const sections = ["# Action Items", ""];
   sections.push(...renderSynthesisItems(synthesisItems, speakerMapping));
-  sections.push(...renderAssignedTasks(assignedTasks, speakerMapping));
+  sections.push(...renderAssignedTasks(dedupedTasks, speakerMapping));
   while (sections[sections.length - 1] === "") sections.pop();
   return sections.join("\n");
 }
 
-// src/output/timeline.ts
-function toPercent(seconds, duration) {
-  if (duration <= 0) return "0";
-  const pct = Math.min(100, Math.max(0, seconds / duration * 100));
-  return pct.toFixed(3);
-}
-function escapeHtml(str) {
-  return str.replace(/&/g, "&amp;").replace(/</g, "&lt;").replace(/>/g, "&gt;").replace(/"/g, "&quot;").replace(/'/g, "&#39;");
-}
-function collectMarkers(pipelineResult, duration, speakerMapping) {
-  const markers = [];
-  const SPEECH_WINDOW_SECONDS = 30;
-  for (const segment of pipelineResult.segments) {
-    if (segment.pass1 != null) {
-      let windowStart = -1;
-      let windowLabel = "";
-      let windowDetail = "";
-      for (const entry of segment.pass1.transcript_entries) {
-        const seconds = parseTimestamp(entry.timestamp);
-        if (seconds > duration && duration > 0) continue;
-        if (windowStart < 0 || seconds - windowStart >= SPEECH_WINDOW_SECONDS) {
-          if (windowStart >= 0) {
-            markers.push({ seconds: windowStart, label: windowLabel, lane: "speech", detail: windowDetail });
-          }
-          windowStart = seconds;
-          windowLabel = applySpeakerMapping(entry.speaker || "Speech", speakerMapping);
-          windowDetail = entry.text.slice(0, 80) + (entry.text.length > 80 ? "\u2026" : "");
-        }
-      }
-      if (windowStart >= 0) {
-        markers.push({ seconds: windowStart, label: windowLabel, lane: "speech", detail: windowDetail });
-      }
-    }
-    if (segment.pass2 != null) {
-      for (const block of segment.pass2.code_blocks) {
-        const seconds = parseTimestamp(block.timestamp);
-        if (seconds > duration && duration > 0) continue;
-        markers.push({
-          seconds,
-          lane: "code",
-          label: block.filename,
-          detail: block.language + (block.change_type ? ` \xB7 ${block.change_type}` : "")
-        });
-      }
-      for (const note2 of segment.pass2.visual_notes) {
-        const seconds = parseTimestamp(note2.timestamp);
-        if (seconds > duration && duration > 0) continue;
-        markers.push({
-          seconds,
-          lane: "visual",
-          label: note2.visual_type,
-          detail: note2.description.slice(0, 80) + (note2.description.length > 80 ? "\u2026" : "")
-        });
-      }
-    }
-  }
-  const topics = pipelineResult.synthesisResult?.topics ?? [];
-  for (const topic of topics) {
-    for (const ts of topic.timestamps) {
-      const seconds = parseTimestamp(ts);
-      if (seconds > duration && duration > 0) continue;
-      markers.push({
-        seconds,
-        lane: "topic",
-        label: topic.title,
-        detail: topic.summary?.slice(0, 80) ?? ""
-      });
-    }
-  }
-  return markers;
-}
-function renderMarker(m, duration) {
-  const left = toPercent(m.seconds, duration);
-  const time = formatTime(m.seconds);
-  const tooltipRaw = `${time} \u2014 ${m.label}${m.detail ? ": " + m.detail : ""}`;
-  const tooltipAttr = escapeHtml(tooltipRaw);
-  return `<div class="marker marker-${m.lane}" style="left:${left}%" title="${tooltipAttr}" aria-label="${tooltipAttr}"></div>`;
-}
-function renderLane(laneId, laneLabel, markers, duration) {
-  const laneMarkers = markers.filter((m) => m.lane === laneId);
-  const renderedMarkers = laneMarkers.map((m) => renderMarker(m, duration)).join("\n        ");
-  return `
-    <div class="lane">
-      <div class="lane-label">${laneLabel}</div>
-      <div class="lane-track" role="region" aria-label="${laneLabel} lane">
-        ${renderedMarkers}
-      </div>
-    </div>`;
-}
-function buildTimeAxis(duration) {
-  if (duration <= 0) return "";
-  const intervals = [30, 60, 120, 300, 600, 900, 1800, 3600];
-  const targetTicks = 10;
-  const ideal = duration / targetTicks;
-  const interval = intervals.find((i) => i >= ideal) ?? intervals[intervals.length - 1] ?? 3600;
-  const ticks = [];
-  for (let t = 0; t <= duration; t += interval) {
-    const left = toPercent(t, duration);
-    const label = formatTime(t);
-    ticks.push(`<div class="tick" style="left:${left}%"><span>${label}</span></div>`);
-  }
-  return ticks.join("\n        ");
-}
-function generateTimeline(params) {
-  const { pipelineResult, duration, speakerMapping } = params;
-  const markers = collectMarkers(pipelineResult, duration, speakerMapping);
-  const effectiveDuration = duration > 0 ? duration : 1;
-  const speechLane = renderLane("speech", "Speech", markers, effectiveDuration);
-  const codeLane = renderLane("code", "Code", markers, effectiveDuration);
-  const visualLane = renderLane("visual", "Slides / Visuals", markers, effectiveDuration);
-  const topicLane = renderLane("topic", "Key Moments", markers, effectiveDuration);
-  const timeAxis = buildTimeAxis(effectiveDuration);
-  return `<!DOCTYPE html>
-<html lang="en">
-<head>
-  <meta charset="UTF-8">
-  <meta name="viewport" content="width=device-width, initial-scale=1.0">
-  <title>Video Timeline</title>
-  <style>
-    /* ------------------------------------------------------------------ */
-    /* Reset + base                                                        */
-    /* ------------------------------------------------------------------ */
-    *, *::before, *::after { box-sizing: border-box; margin: 0; padding: 0; }
-
-    :root {
-      --bg: #f9fafb;
-      --surface: #ffffff;
-      --border: #e5e7eb;
-      --text: #111827;
-      --text-muted: #6b7280;
-      --shadow: 0 1px 3px rgba(0,0,0,.1);
-
-      --speech-color: #3b82f6;
-      --code-color: #22c55e;
-      --visual-color: #a855f7;
-      --topic-color: #eab308;
-
-      --lane-h: 36px;
-      --label-w: 120px;
-    }
-
-    @media (prefers-color-scheme: dark) {
-      :root {
-        --bg: #0f172a;
-        --surface: #1e293b;
-        --border: #334155;
-        --text: #f1f5f9;
-        --text-muted: #94a3b8;
-        --shadow: 0 1px 3px rgba(0,0,0,.4);
-      }
-    }
-
-    html, body {
-      height: 100%;
-      background: var(--bg);
-      color: var(--text);
-      font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, sans-serif;
-      font-size: 14px;
-      line-height: 1.5;
-    }
-
-    /* ------------------------------------------------------------------ */
-    /* Layout                                                             */
-    /* ------------------------------------------------------------------ */
-    .page {
-      max-width: 100%;
-      padding: 1.5rem 1rem;
-    }
-
-    h1 {
-      font-size: 1.25rem;
-      font-weight: 700;
-      margin-bottom: 1rem;
-    }
-
-    .legend {
-      display: flex;
-      flex-wrap: wrap;
-      gap: .5rem 1rem;
-      margin-bottom: 1.25rem;
-    }
-
-    .legend-item {
-      display: flex;
-      align-items: center;
-      gap: .4rem;
-      font-size: .8125rem;
-      color: var(--text-muted);
-    }
-
-    .legend-dot {
-      width: 10px;
-      height: 10px;
-      border-radius: 50%;
-      flex-shrink: 0;
-    }
-
-    .legend-dot.speech  { background: var(--speech-color); }
-    .legend-dot.code    { background: var(--code-color); }
-    .legend-dot.visual  { background: var(--visual-color); }
-    .legend-dot.topic   { background: var(--topic-color); }
-
-    /* ------------------------------------------------------------------ */
-    /* Scroll container                                                   */
-    /* ------------------------------------------------------------------ */
-    .scroll-wrapper {
-      overflow-x: auto;
-      -webkit-overflow-scrolling: touch;
-      border: 1px solid var(--border);
-      border-radius: .5rem;
-      background: var(--surface);
-      box-shadow: var(--shadow);
-    }
-
-    .timeline {
-      min-width: 640px;
-      padding: 1rem;
-    }
-
-    /* ------------------------------------------------------------------ */
-    /* Lanes                                                              */
-    /* ------------------------------------------------------------------ */
-    .lane {
-      display: flex;
-      align-items: center;
-      margin-bottom: .5rem;
-    }
-
-    .lane-label {
-      width: var(--label-w);
-      flex-shrink: 0;
-      font-size: .75rem;
-      font-weight: 600;
-      color: var(--text-muted);
-      text-transform: uppercase;
-      letter-spacing: .04em;
-      padding-right: .75rem;
-    }
-
-    .lane-track {
-      flex: 1;
-      height: var(--lane-h);
-      position: relative;
-      background: var(--bg);
-      border-radius: .25rem;
-      border: 1px solid var(--border);
-    }
-
-    /* ------------------------------------------------------------------ */
-    /* Markers                                                            */
-    /* ------------------------------------------------------------------ */
-    .marker {
-      position: absolute;
-      top: 50%;
-      transform: translate(-50%, -50%);
-      width: 10px;
-      height: 10px;
-      border-radius: 50%;
-      cursor: pointer;
-      transition: transform .15s ease, box-shadow .15s ease;
-      outline: none;
-    }
-
-    .marker:hover, .marker:focus {
-      transform: translate(-50%, -50%) scale(1.6);
-      box-shadow: 0 0 0 2px var(--surface), 0 0 0 4px currentColor;
-      z-index: 10;
-    }
-
-    .marker-speech { background: var(--speech-color); color: var(--speech-color); }
-    .marker-code   { background: var(--code-color);   color: var(--code-color); }
-    .marker-visual { background: var(--visual-color); color: var(--visual-color); }
-    .marker-topic  { background: var(--topic-color);  color: var(--topic-color); }
-
-    /* ------------------------------------------------------------------ */
-    /* Time axis                                                          */
-    /* ------------------------------------------------------------------ */
-    .time-axis {
-      display: flex;
-      align-items: center;
-      margin-bottom: .25rem;
-    }
-
-    .axis-spacer {
-      width: var(--label-w);
-      flex-shrink: 0;
-    }
-
-    .axis-track {
-      flex: 1;
-      height: 20px;
-      position: relative;
-    }
-
-    .tick {
-      position: absolute;
-      top: 0;
-      transform: translateX(-50%);
-    }
-
-    .tick span {
-      font-size: .6875rem;
-      color: var(--text-muted);
-      white-space: nowrap;
-    }
-
-    /* ------------------------------------------------------------------ */
-    /* Tooltip (vanilla JS)                                               */
-    /* ------------------------------------------------------------------ */
-    #tooltip {
-      position: fixed;
-      background: var(--surface);
-      border: 1px solid var(--border);
-      border-radius: .375rem;
-      padding: .375rem .625rem;
-      font-size: .8125rem;
-      color: var(--text);
-      pointer-events: none;
-      z-index: 999;
-      box-shadow: 0 4px 12px rgba(0,0,0,.15);
-      max-width: 280px;
-      word-break: break-word;
-      display: none;
-    }
-
-    /* ------------------------------------------------------------------ */
-    /* Responsive                                                         */
-    /* ------------------------------------------------------------------ */
-    @media (max-width: 480px) {
-      :root { --label-w: 72px; --lane-h: 32px; }
-      .legend { gap: .35rem .75rem; }
-      h1 { font-size: 1.1rem; }
-    }
-  </style>
-</head>
-<body>
-  <div class="page">
-    <h1>Video Timeline</h1>
-
-    <div class="legend" aria-label="Lane colour legend">
-      <span class="legend-item"><span class="legend-dot speech"></span>Speech</span>
-      <span class="legend-item"><span class="legend-dot code"></span>Code</span>
-      <span class="legend-item"><span class="legend-dot visual"></span>Slides / Visuals</span>
-      <span class="legend-item"><span class="legend-dot topic"></span>Key Moments</span>
-    </div>
-
-    <div class="scroll-wrapper">
-      <div class="timeline">
-        <div class="time-axis">
-          <div class="axis-spacer"></div>
-          <div class="axis-track">
-            ${timeAxis}
-          </div>
-        </div>
-        ${speechLane}
-        ${codeLane}
-        ${visualLane}
-        ${topicLane}
-      </div>
-    </div>
-  </div>
-
-  <div id="tooltip" role="tooltip"></div>
-
-  <script>
-    (function () {
-      var tooltip = document.getElementById('tooltip');
-      var markers = document.querySelectorAll('.marker');
-
-      function showTooltip(el, x, y) {
-        var text = el.getAttribute('title') || el.getAttribute('aria-label') || '';
-        if (!text) return;
-        tooltip.textContent = text;
-        tooltip.style.display = 'block';
-        positionTooltip(x, y);
-      }
-
-      function positionTooltip(x, y) {
-        var tw = tooltip.offsetWidth;
-        var th = tooltip.offsetHeight;
-        var vw = window.innerWidth;
-        var vh = window.innerHeight;
-        var left = x + 12;
-        var top = y - th / 2;
-        if (left + tw > vw - 8) left = x - tw - 12;
-        if (top < 8) top = 8;
-        if (top + th > vh - 8) top = vh - th - 8;
-        tooltip.style.left = left + 'px';
-        tooltip.style.top = top + 'px';
-      }
-
-      function hideTooltip() {
-        tooltip.style.display = 'none';
-      }
-
-      markers.forEach(function (m) {
-        m.addEventListener('mouseenter', function (e) {
-          showTooltip(m, e.clientX, e.clientY);
-        });
-        m.addEventListener('mousemove', function (e) {
-          positionTooltip(e.clientX, e.clientY);
-        });
-        m.addEventListener('mouseleave', hideTooltip);
-        m.addEventListener('focus', function () {
-          var rect = m.getBoundingClientRect();
-          showTooltip(m, rect.right, rect.top + rect.height / 2);
-        });
-        m.addEventListener('blur', hideTooltip);
-      });
-    })();
-  </script>
-</body>
-</html>`;
-}
-
 // src/output/metadata.ts
 function writeMetadata(params) {
   const { title, source, duration, model, processingTimeMs, filesGenerated, pipelineResult, speakerMapping, declinedMerges } = params;
@@ -4722,12 +4439,10 @@ function resolveFilesToGenerate(params) {
   }
   if (synthesisResult != null || hasPass3d) optional.add("notes.md");
   if (peopleExtraction != null) optional.add("people.md");
-  const hasPass1 = segments.some((s) => s.pass1 != null);
-  if (hasPass1 || hasPass2) optional.add("timeline.html");
   return optional;
 }
 async function generateOutput(params) {
-  const { pipelineResult, outputDir, videoTitle, source, duration, model, processingTimeMs, speakerMapping, declinedMerges } = params;
+  const { pipelineResult, outputDir, videoTitle, source, duration, model, processingTimeMs, channelAuthor, speakerMapping, declinedMerges } = params;
   const slug = slugify(videoTitle);
   const finalOutputDir = join3(outputDir, slug);
   await mkdir(finalOutputDir, { recursive: true });
@@ -4832,14 +4547,6 @@ async function generateOutput(params) {
       errors.push(`action-items.md: ${String(err)}`);
     }
   }
-  if (filesToGenerate.has("timeline.html")) {
-    try {
-      const content = generateTimeline({ pipelineResult, duration, speakerMapping: expandedMapping });
-      await writeOutputFile("timeline.html", content);
-    } catch (err) {
-      errors.push(`timeline.html: ${String(err)}`);
-    }
-  }
   try {
     const rawFiles = writeRawOutput(pipelineResult);
     await mkdir(join3(finalOutputDir, "raw"), { recursive: true });
@@ -4869,7 +4576,7 @@ async function generateOutput(params) {
     errors.push(`metadata.json: ${String(err)}`);
   }
   try {
-    const content = writeGuide({ title: videoTitle, source, duration, pipelineResult, filesGenerated, speakerMapping: expandedMapping });
+    const content = writeGuide({ title: videoTitle, source, duration, pipelineResult, filesGenerated, speakerMapping: expandedMapping, channelAuthor });
     await writeOutputFile("guide.md", content);
   } catch (err) {
     errors.push(`guide.md: ${String(err)}`);
@@ -4982,14 +4689,6 @@ async function reRenderWithSpeakerMapping(params) {
       errors.push(`action-items.md: ${String(err)}`);
     }
   }
-  if (filesToReRender.has("timeline.html")) {
-    try {
-      const content = generateTimeline({ pipelineResult, duration, speakerMapping: expandedMapping });
-      await writeOutputFile("timeline.html", content);
-    } catch (err) {
-      errors.push(`timeline.html: ${String(err)}`);
-    }
-  }
   if (filesToReRender.has("guide.md")) {
     try {
       const content = writeGuide({ title: videoTitle, source, duration, pipelineResult, filesGenerated, speakerMapping: expandedMapping });
@@ -5168,6 +4867,7 @@ async function runDistill(args) {
   let duration;
   let videoTitle;
   let uploadedFileNames = [];
+  let ytAuthor;
   if (resolved.type === "youtube") {
     const result = await handleYouTube(resolved.value, client);
     fileUri = result.fileUri;
@@ -5184,8 +4884,14 @@ async function runDistill(args) {
     if (result.uploadedFileName != null) {
       uploadedFileNames = [result.uploadedFileName];
     }
-    const videoId = extractVideoId(resolved.value);
-    videoTitle = videoId != null ? `youtube-${videoId}` : resolved.value;
+    try {
+      const meta = await fetchYouTubeMetadata(resolved.value);
+      videoTitle = meta.title;
+      ytAuthor = meta.author;
+    } catch {
+      const videoId = extractVideoId(resolved.value);
+      videoTitle = videoId != null ? `youtube-${videoId}` : resolved.value;
+    }
   } else {
     const result = await handleLocalFile(resolved.value, client);
     fileUri = result.fileUri;
@@ -5230,6 +4936,7 @@ async function runDistill(args) {
     model,
     context,
     lang: args.lang,
+    channelAuthor: ytAuthor,
     rateLimiter,
     onProgress: (status) => {
       progress2.update(status);
@@ -5253,7 +4960,8 @@ async function runDistill(args) {
     source: rawInput,
     duration,
     model,
-    processingTimeMs: elapsedMs
+    processingTimeMs: elapsedMs,
+    channelAuthor: ytAuthor
   });
   const elapsedSecs = Math.round(elapsedMs / 1e3);
   const elapsedMins = Math.floor(elapsedSecs / 60);
@@ -5293,6 +5001,7 @@ async function analyzeVideo(input, context, lang) {
   let mimeType;
   let duration;
   let videoTitle;
+  let ytAuthor;
   if (resolved.type === "youtube") {
     const result = await handleYouTube(resolved.value, client);
     fileUri = result.fileUri;
@@ -5307,8 +5016,14 @@ async function analyzeVideo(input, context, lang) {
 `);
       duration = 600;
     }
-    const videoId = extractVideoId(resolved.value);
-    videoTitle = videoId != null ? `youtube-${videoId}` : resolved.value;
+    try {
+      const meta = await fetchYouTubeMetadata(resolved.value);
+      videoTitle = meta.title;
+      ytAuthor = meta.author;
+    } catch {
+      const videoId = extractVideoId(resolved.value);
+      videoTitle = videoId != null ? `youtube-${videoId}` : resolved.value;
+    }
   } else {
     const result = await handleLocalFile(resolved.value, client);
     fileUri = result.fileUri;
@@ -5324,6 +5039,7 @@ async function analyzeVideo(input, context, lang) {
   const slug = slugify(videoTitle);
   const finalOutputDir = `${outputDir}/${slug}`;
   const rateLimiter = new RateLimiter();
+  const startTime = Date.now();
   const pipelineResult = await runPipeline({
     client,
     fileUri,
@@ -5332,8 +5048,10 @@ async function analyzeVideo(input, context, lang) {
     model,
     context,
     lang,
+    channelAuthor: ytAuthor,
     rateLimiter
   });
+  const elapsedMs = Date.now() - startTime;
   await generateOutput({
     pipelineResult,
     outputDir,
@@ -5341,7 +5059,8 @@ async function analyzeVideo(input, context, lang) {
     source: input,
     duration,
     model,
-    processingTimeMs: 0
+    processingTimeMs: elapsedMs,
+    channelAuthor: ytAuthor
   });
   let summary = "Analysis complete.";
   const synthesisPath = join5(finalOutputDir, "raw", "synthesis.json");
@@ -5902,7 +5621,7 @@ async function run2(args) {
 }
 
 // src/cli/index.ts
-var version = "0.6.2";
+var version = "0.6.4";
 var DEFAULT_OUTPUT2 = "./vidistill-output/";
 var SUBCOMMANDS = {
   mcp: run,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "vidistill",
-  "version": "0.6.2",
+  "version": "0.6.4",
   "description": "Video intelligence distiller — extract structured notes, transcripts, and insights from any video using Gemini",
   "type": "module",
   "license": "MIT",