Skip to content

YC (S26) | Open Computer History | Continuously record your company computer work, map your workflows, help you find work worth automating, and power your agents' context

fix(meetings): keep mic and meeting-audio speaker labels apart in transcript text (SCR-705) (#7613)

Ezra Ellette13d960d27f7c
diff --git a/apps/screenpipe-app-tauri/components/meeting-notes/transcript-panel.tsx b/apps/screenpipe-app-tauri/components/meeting-notes/transcript-panel.tsx
index 9aedf9f8e..f9ab92849 100644
--- a/apps/screenpipe-app-tauri/components/meeting-notes/transcript-panel.tsx
+++ b/apps/screenpipe-app-tauri/components/meeting-notes/transcript-panel.tsx
@@ -29,6 +29,7 @@ import { usePlatform } from "@/lib/hooks/use-platform";
import type { LiveCaptureState } from "@/lib/utils/live-capture-state";
import {
fetchMeetingAudio,
+ sourceQualifiedSpeakerName,
type MeetingAudioChunk,
} from "@/lib/utils/meeting-context";
import type { MeetingRecord } from "@/lib/utils/meeting-format";
@@ -401,11 +402,7 @@ function compareBlocks(a: SpeakerBlock, b: SpeakerBlock): number {
}
function displaySpeakerName(block: SpeakerBlock): string {
- if (!/^speaker(?:[ _]\d+)?$/i.test(block.speakerName.trim())) return block.speakerName;
- // A provider's speaker numbers restart on each stream. Avoid presenting
- // unrelated input/output labels as the same identified person.
- const source = block.captureDirection === "input" ? "Microphone" : block.captureDirection === "output" ? "Meeting audio" : "Unidentified";
- return `${source} · ${block.speakerName}`;
+ return sourceQualifiedSpeakerName(block.speakerName, block.captureDirection);
}
function groupBySpeaker(chunks: MeetingAudioChunk[]): SpeakerBlock[] {
@@ -1212,7 +1209,7 @@ export function TranscriptPanel({
aria-atomic="true"
>
{latestFinalLiveBlock
- ? `${latestFinalLiveBlock.speakerName}: ${latestFinalLiveBlock.text}`
+ ? `${displaySpeakerName(latestFinalLiveBlock)}: ${latestFinalLiveBlock.text}`
: ""}
</span>
</div>
diff --git a/apps/screenpipe-app-tauri/lib/utils/meeting-context.test.ts b/apps/screenpipe-app-tauri/lib/utils/meeting-context.test.ts
index 5993f64b8..3251a67a0 100644
--- a/apps/screenpipe-app-tauri/lib/utils/meeting-context.test.ts
+++ b/apps/screenpipe-app-tauri/lib/utils/meeting-context.test.ts
@@ -10,6 +10,7 @@ import {
extractImageDataUrlsFromMarkdown,
extractPipePromptBody,
mergeMeetingAudioChunks,
+ renderMeetingTranscript,
type MeetingAudioChunk,
type MeetingContext,
} from "./meeting-context";
@@ -213,6 +214,20 @@ describe("mergeMeetingAudioChunks", () => {
});
});
+describe("renderMeetingTranscript speaker labels", () => {
+ it("keeps per-stream provider labels apart for mic and meeting audio", () => {
+ const remote = chunk({ speakerName: "speaker 1", source: "live", transcription: "remote turn", timestamp: "2026-06-04T15:00:00Z" });
+ const local = chunk({ audioChunkId: 2, speakerName: "speaker 1", source: "live", deviceType: "input", isInput: true, transcription: "local turn", timestamp: "2026-06-04T15:00:03Z" });
+ const named = chunk({ audioChunkId: 3, speakerName: "Alice", source: "live", deviceType: "input", isInput: true, transcription: "named turn", timestamp: "2026-06-04T15:00:30Z" });
+
+ const lines = renderMeetingTranscript([remote, local, named]).split("\n");
+
+ expect(lines).toHaveLength(3);
+ expect(lines[0]).toContain("[Meeting audio · speaker 1] remote turn");
+ expect(lines[1]).toContain("[Microphone · speaker 1] local turn");
+ expect(lines[2]).toContain("[Alice] named turn");
+ });
+});
describe("cross-device speech preservation", () => {
it.each([
diff --git a/apps/screenpipe-app-tauri/lib/utils/meeting-context.ts b/apps/screenpipe-app-tauri/lib/utils/meeting-context.ts
index e3b7b28cc..dc89bc852 100644
--- a/apps/screenpipe-app-tauri/lib/utils/meeting-context.ts
+++ b/apps/screenpipe-app-tauri/lib/utils/meeting-context.ts
@@ -129,6 +129,17 @@ function sortAudioChunks(chunks: MeetingAudioChunk[]): MeetingAudioChunk[] {
});
}
+/** Provider speaker numbers restart on each stream, so qualify generic labels by source. */
+export function sourceQualifiedSpeakerName(
+ speakerName: string,
+ deviceType: string | undefined,
+): string {
+ if (!/^speaker(?:[ _]\d+)?$/i.test(speakerName.trim())) return speakerName;
+ const direction = deviceType?.trim().toLowerCase();
+ const source = direction === "input" ? "Microphone" : direction === "output" ? "Meeting audio" : "Unidentified";
+ return `${source} · ${speakerName}`;
+}
+
interface SpeakerTurn {
speakerName: string;
/** Timestamp of the first chunk in this turn — used for display. */
@@ -154,7 +165,7 @@ function groupConsecutiveSpeakerTurns(
const MAX_GAP_MS = 10_000;
const turns: SpeakerTurn[] = [];
for (const c of chunks) {
- const speakerName = c.speakerName ?? "";
+ const speakerName = sourceQualifiedSpeakerName(c.speakerName ?? "", c.deviceType);
const text = c.transcription.replace(/\s+/g, " ").trim();
if (!text) continue;
const last = turns[turns.length - 1];