diff --git a/src/audio/VoiceRecording.ts b/src/audio/VoiceRecording.ts
index 1c6c8b40d2..d0b34493d8 100644
--- a/src/audio/VoiceRecording.ts
+++ b/src/audio/VoiceRecording.ts
@@ -37,7 +37,7 @@ import mxRecorderWorkletPath from "./RecorderWorklet";
 const CHANNELS = 1; // stereo isn't important
 export const SAMPLE_RATE = 48000; // 48khz is what WebRTC uses. 12khz is where we lose quality.
 const BITRATE = 24000; // 24kbps is pretty high quality for our use case in opus.
-const TARGET_MAX_LENGTH = 120; // 2 minutes in seconds. Somewhat arbitrary, though longer == larger files.
+const TARGET_MAX_LENGTH = 900; // 15 minutes in seconds. Somewhat arbitrary, though longer == larger files.
 const TARGET_WARN_TIME_LEFT = 10; // 10 seconds, also somewhat arbitrary.
 
 export const RECORDING_PLAYBACK_SAMPLES = 44;
@@ -229,8 +229,8 @@ export class VoiceRecording extends EventEmitter implements IDestroyable {
         // go horribly over the limit. We also emit a warning state if needed.
         //
         // We use the recorder's perspective of time to make sure we don't cut off the last
-        // frame of audio, otherwise we end up with a 1:59 clip (119.68 seconds). This extra
-        // safety can allow us to overshoot the target a bit, but at least when we say 2min
+        // frame of audio, otherwise we end up with a 14:59 clip (899.68 seconds). This extra
+        // safety can allow us to overshoot the target a bit, but at least when we say 15min
         // maximum we actually mean it.
         //
         // In testing, recorder time and worker time lag by about 400ms, which is roughly the