vidistill 0.5.2 → 0.5.4

package/dist/index.js

@@ -1597,6 +1597,10 @@ function applySpeakerMapping(label, mapping) {
   for (const [key, value] of Object.entries(mapping)) {
     if (key.toLowerCase() === lower) return value;
   }
+  for (const [key, value] of Object.entries(mapping)) {
+    const keyParen = key.match(/\(([^)]+)\)/);
+    if (keyParen && keyParen[1].trim().toLowerCase() === lower) return value;
+  }
   return label;
 }
 function replaceNamesInText(text4, mapping) {
@@ -1604,14 +1608,22 @@ function replaceNamesInText(text4, mapping) {
   const entries = Object.entries(mapping).filter(([key, value]) => key !== value && !/^SPEAKER_\d+$/.test(key)).sort((a, b) => b[0].length - a[0].length);
   if (entries.length === 0) return text4;
   let result = text4;
+  const placeholders = /* @__PURE__ */ new Map();
+  let idx = 0;
   for (const [key, value] of entries) {
+    const placeholder = `\0PH${idx}\0`;
+    placeholders.set(placeholder, value);
     const escaped = key.replace(/[.*+?^${}()|[\]\\]/g, "\\$&");
     const re = new RegExp(`\\b${escaped}\\b`, "g");
-    result = result.replace(re, value);
+    result = result.replace(re, placeholder);
+    idx++;
+  }
+  for (const [placeholder, value] of placeholders) {
+    result = result.replaceAll(placeholder, value);
   }
   return result;
 }
-function buildExpandedMapping(segments, speakerMapping) {
+function buildExpandedMapping(segments, speakerMapping, peopleExtraction) {
   const expanded = { ...speakerMapping };
   for (const seg of segments) {
     if (seg.pass1 == null) continue;
@@ -1640,6 +1652,19 @@ function buildExpandedMapping(segments, speakerMapping) {
       }
     }
   }
+  if (peopleExtraction?.participants != null) {
+    for (const p of peopleExtraction.participants) {
+      if (!p.name || expanded[p.name] != null) continue;
+      for (const [key, value] of Object.entries(expanded)) {
+        if (/^SPEAKER_\d+$/.test(key)) continue;
+        if (key === value) continue;
+        if (p.name !== key && p.name.includes(key)) {
+          expanded[p.name] = value;
+          break;
+        }
+      }
+    }
+  }
   return expanded;
 }
 async function readJsonFile(filePath) {
@@ -2009,6 +2034,36 @@ async function runLinkConsensus(params) {
 
 // src/core/transcript-consensus.ts
 var ALIGN_WINDOW_S = 3;
+var DEDUP_WINDOW_S = 10;
+function isNearDuplicate(a, b) {
+  const delta = Math.abs(parseTimestamp(a.timestamp) - parseTimestamp(b.timestamp));
+  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;
+}
+function deduplicateEntries(entries) {
+  if (entries.length <= 1) return entries;
+  const result = [entries[0]];
+  for (let i = 1; i < entries.length; i++) {
+    const curr = entries[i];
+    let isDup = false;
+    for (let j = result.length - 1; j >= Math.max(0, result.length - 3); j--) {
+      if (isNearDuplicate(curr, result[j])) {
+        if (curr.text.length > result[j].text.length) {
+          result[j] = curr;
+        }
+        isDup = true;
+        break;
+      }
+    }
+    if (!isDup) {
+      result.push(curr);
+    }
+  }
+  return result;
+}
 function selectBestText(texts) {
   if (texts.length === 1) return texts[0];
   const referenceText = texts.join(" ");
@@ -2024,7 +2079,12 @@ function selectBestText(texts) {
   return bestText;
 }
 function mergeTranscriptRuns(runs) {
-  if (runs.length === 1) return runs[0];
+  if (runs.length === 1) {
+    return {
+      ...runs[0],
+      transcript_entries: deduplicateEntries(runs[0].transcript_entries)
+    };
+  }
   const referenceRun = runs.reduce(
     (best, run3) => run3.transcript_entries.length > best.transcript_entries.length ? run3 : best
   );
@@ -2071,7 +2131,7 @@ function mergeTranscriptRuns(runs) {
   return {
     segment_index: referenceRun.segment_index,
     time_range: referenceRun.time_range,
-    transcript_entries: mergedEntries
+    transcript_entries: deduplicateEntries(mergedEntries)
   };
 }
 function mergeSpeakerSummaries(summaries) {
@@ -3482,37 +3542,30 @@ function writeGuide(params) {
 }
 
 // src/output/transcript.ts
-var PAUSE_THRESHOLD_SECONDS = 1.5;
-function applyEmphasis(text4, emphasisWords) {
-  if (emphasisWords == null || emphasisWords.length === 0) return text4;
-  let result = text4;
-  for (const word of emphasisWords) {
-    const escaped = word.replace(/[.*+?^${}()|[\]\\]/g, "\\$&");
-    const re = new RegExp(`(?<![\\w*])${escaped}(?![\\w*])`, "gi");
-    result = result.replace(re, `**$&**`);
-  }
-  return result;
-}
 function renderEntry(entry, speakerMapping) {
-  const emphasized = applyEmphasis(entry.text, entry.emphasis_words);
-  const pause = entry.pause_after_seconds != null && entry.pause_after_seconds >= PAUSE_THRESHOLD_SECONDS ? ` _(pause ${entry.pause_after_seconds.toFixed(1)}s)_` : "";
   const speaker = applySpeakerMapping(entry.speaker, speakerMapping);
-  return `**[${entry.timestamp}]** **${speaker}:** ${emphasized}${pause}`;
+  return `**[${entry.timestamp}]** **${speaker}:** ${entry.text}`;
+}
+function parseEndTime(timeRange) {
+  const parts = timeRange.split("-");
+  if (parts.length < 2) return Infinity;
+  const endStr = parts[parts.length - 1].trim();
+  const seconds = parseTimestamp(endStr);
+  return seconds > 0 ? seconds : Infinity;
 }
 function renderPass1(pass1, speakerMapping) {
   const lines = [];
   lines.push(`### Segment ${pass1.segment_index + 1} \u2014 ${pass1.time_range}`);
   lines.push("");
-  if (pass1.speaker_summary.length > 0) {
-    lines.push(
-      "_Speakers: " + pass1.speaker_summary.map((s) => `${applySpeakerMapping(s.speaker_id, speakerMapping)} (${s.description})`).join(", ") + "_"
-    );
-    lines.push("");
-  }
-  if (pass1.transcript_entries.length === 0) {
+  const endTime = parseEndTime(pass1.time_range);
+  const validEntries = pass1.transcript_entries.filter((entry) => {
+    const entryTime = parseTimestamp(entry.timestamp);
+    return entryTime <= endTime;
+  });
+  if (validEntries.length === 0) {
     lines.push("_No transcript entries for this segment._");
   } else {
-    for (const entry of pass1.transcript_entries) {
+    for (const entry of validEntries) {
       lines.push(renderEntry(entry, speakerMapping));
     }
   }
@@ -4681,7 +4734,7 @@ async function generateOutput(params) {
   const filesGenerated = [];
   const errors = [];
   const filesToGenerate = resolveFilesToGenerate(params);
-  const expandedMapping = speakerMapping ? buildExpandedMapping(pipelineResult.segments, speakerMapping) : void 0;
+  const expandedMapping = speakerMapping ? buildExpandedMapping(pipelineResult.segments, speakerMapping, pipelineResult.peopleExtraction) : void 0;
   async function writeOutputFile(filename, content) {
     const fullPath = join3(finalOutputDir, filename);
     const dir = dirname(fullPath);
@@ -4875,7 +4928,7 @@ async function reRenderWithSpeakerMapping(params) {
     await writeFile(fullPath, content, "utf8");
     filesWritten.push(filename);
   }
-  const expandedMapping = buildExpandedMapping(pipelineResult.segments, speakerMapping);
+  const expandedMapping = buildExpandedMapping(pipelineResult.segments, speakerMapping, pipelineResult.peopleExtraction);
   const filesToReRender = new Set(filesGenerated.filter((f) => !f.startsWith("raw/")));
   if (filesToReRender.has("transcript.md")) {
     try {
@@ -5638,7 +5691,7 @@ async function run2(args) {
 }
 
 // src/cli/index.ts
-var version = "0.5.2";
+var version = "0.5.4";
 var DEFAULT_OUTPUT = "./vidistill-output/";
 var SUBCOMMANDS = {
   mcp: run,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "vidistill",
-  "version": "0.5.2",
+  "version": "0.5.4",
   "description": "Video intelligence distiller — extract structured notes, transcripts, and insights from any video using Gemini",
   "type": "module",
   "license": "MIT",