fix(cli): assign w{index} ids when loading JSON transcripts

packages/cli/src/whisper/normalize.test.ts

@@ -103,8 +103,8 @@ describe("loadTranscript", () => {
     const { words, format } = loadTranscript(path);
     expect(format).toBe("whisper-cpp");
     expect(words).toEqual([
-      { text: "Hello,", start: 0, end: 0.55 },
-      { text: "world.", start: 0.6, end: 1.25 },
+      { text: "Hello,", start: 0, end: 0.55, id: "w0" },
+      { text: "world.", start: 0.6, end: 1.25, id: "w1" },
     ]);
   });
 
@@ -142,8 +142,8 @@ describe("loadTranscript", () => {
     const { words, format } = loadTranscript(path);
     expect(format).toBe("openai");
     expect(words).toEqual([
-      { text: "Hello", start: 0, end: 0.5 },
-      { text: "world", start: 0.6, end: 1.2 },
+      { text: "Hello", start: 0, end: 0.5, id: "w0" },
+      { text: "world", start: 0.6, end: 1.2, id: "w1" },
     ]);
   });
 
@@ -205,7 +205,7 @@ Short format
     expect(words[0]?.text).toBe("Bold and italic");
   });
 
-  it("passes through normalized word arrays", () => {
+  it("assigns w{index} ids to normalized word arrays", () => {
     const input = [
       { text: "Hello", start: 0.0, end: 0.5 },
       { text: "world", start: 0.6, end: 1.2 },
@@ -214,10 +214,21 @@ Short format
     const { words, format } = loadTranscript(path);
     expect(format).toBe("words-json");
     expect(words).toEqual([
-      { text: "Hello", start: 0, end: 0.5, id: "" },
-      { text: "world", start: 0.6, end: 1.2, id: "" },
+      { text: "Hello", start: 0, end: 0.5, id: "w0" },
+      { text: "world", start: 0.6, end: 1.2, id: "w1" },
     ]);
   });
+
+  it("preserves existing ids and repairs empty-string ids from legacy files", () => {
+    const input = [
+      { text: "Hello", start: 0.0, end: 0.5, id: "keep-me" },
+      { text: "world", start: 0.6, end: 1.2, id: "" },
+      { text: "again", start: 1.3, end: 1.8 },
+    ];
+    const path = tmpFile("legacy.json", JSON.stringify(input));
+    const { words } = loadTranscript(path);
+    expect(words.map((w) => w.id)).toEqual(["keep-me", "w1", "w2"]);
+  });
 });
 
 describe("caption formatting", () => {
@@ -328,9 +339,9 @@ describe("whisper-cpp contraction merging", () => {
     );
     const { words } = loadTranscript(path);
     expect(words).toEqual([
-      { text: "I", start: 0, end: 0.2 },
-      { text: "didn't", start: 0.2, end: 0.7 },
-      { text: "know", start: 0.7, end: 1 },
+      { text: "I", start: 0, end: 0.2, id: "w0" },
+      { text: "didn't", start: 0.2, end: 0.7, id: "w1" },
+      { text: "know", start: 0.7, end: 1, id: "w2" },
     ]);
   });
 

packages/cli/src/whisper/normalize.ts

@@ -462,17 +462,21 @@ export function loadTranscript(filePath: string): { words: Word[]; format: Trans
   const parsed = JSON.parse(content);
   const format = detectJsonFormat(parsed);
 
-  const words =
+  // JSON parsers never set id — assign w{index} like the srt/vtt branches above
+  // so caption overrides always have a stable key. `||` (not `??`) also repairs
+  // the empty-string ids older CLIs wrote to words-json transcript files.
+  const words = (
     format === "whisper-cpp"
       ? parseWhisperCpp(parsed)
       : format === "openai"
         ? parseOpenAI(parsed)
         : (parsed as Word[]).map((w) => ({
-            id: w.id ?? "",
+            id: w.id,
             text: w.text.trim(),
             start: round3(w.start),
             end: round3(w.end),
-          }));
+          }))
+  ).map((w, i) => ({ ...w, id: w.id || `w${i}` }));
 
   return { words, format };
 }