diff --git a/src/main/ipc/audio-handlers.ts b/src/main/ipc/audio-handlers.ts
index 102478e..51d7446 100644
--- a/src/main/ipc/audio-handlers.ts
+++ b/src/main/ipc/audio-handlers.ts
@@ -31,7 +31,15 @@ export function registerAudioHandlers(): void {
   })
 
   ipcMain.handle(IPC_CHANNELS.AUDIO.TEST_DEVICE, async () => {
-    // Phase 1: 스텁
-    return ipcSuccess({ averageLevel: 0, peakLevel: 0, hasAudio: false })
+    try {
+      const service = getAudioCaptureService()
+      const result = await service.testCapture(2000)
+      return ipcSuccess(result)
+    } catch (err) {
+      return ipcError(
+        ErrorCode.AudioDeviceNotFound,
+        `Audio test failed: ${err instanceof Error ? err.message : String(err)}`,
+      )
+    }
   })
 }
diff --git a/src/main/services/AudioCaptureService.ts b/src/main/services/AudioCaptureService.ts
index 729e851..90ea84f 100644
--- a/src/main/services/AudioCaptureService.ts
+++ b/src/main/services/AudioCaptureService.ts
@@ -278,6 +278,61 @@ class AudioCaptureService extends EventEmitter {
     logger.info('AudioCaptureService disposed')
   }
 
+  /**
+   * 지정 시간(ms) 동안 마이크 캡처 후 평균/피크 레벨을 반환한다.
+   */
+  async testCapture(durationMs: number): Promise<{ averageLevel: number; peakLevel: number; hasAudio: boolean }> {
+    const soxExe = getSoxPath()
+    const selectedDeviceId = configGet('selectedDeviceId')
+    const deviceArg = selectedDeviceId && selectedDeviceId !== 'default'
+      ? selectedDeviceId
+      : 'default'
+
+    const soxArgs = process.platform === 'win32'
+      ? ['-t', 'waveaudio', deviceArg, '--no-show-progress',
+         '-r', String(AUDIO_FORMAT.SAMPLE_RATE), '-c', String(AUDIO_FORMAT.CHANNELS),
+         '-b', '16', '-e', 'signed-integer', '-t', 'raw', '-']
+      : ['-d', '--no-show-progress',
+         '-r', String(AUDIO_FORMAT.SAMPLE_RATE), '-c', String(AUDIO_FORMAT.CHANNELS),
+         '-b', '16', '-e', 'signed-integer', '-t', 'raw', '-']
+
+    return new Promise((resolve, reject) => {
+      const proc = spawn(soxExe, soxArgs, { stdio: ['ignore', 'pipe', 'pipe'] })
+      const buffers: Buffer[] = []
+      let peakRms = 0
+
+      proc.stdout?.on('data', (chunk: Buffer) => {
+        buffers.push(chunk)
+        const rms = calculateRMS(chunk)
+        if (rms > peakRms) peakRms = rms
+      })
+
+      proc.on('error', (err) => {
+        reject(new D3ROError(ErrorCode.AudioCaptureStartFailed, `SoX 실행 실패: ${err.message}`))
+      })
+
+      setTimeout(() => {
+        proc.kill()
+      }, durationMs)
+
+      proc.on('close', () => {
+        if (buffers.length === 0) {
+          resolve({ averageLevel: 0, peakLevel: 0, hasAudio: false })
+          return
+        }
+        const combined = Buffer.concat(buffers)
+        const avgRms = calculateRMS(combined)
+        const avgLevel = avgRms > 0 ? Math.min(1.0, Math.pow(avgRms, 0.28)) : 0
+        const peakLevel = peakRms > 0 ? Math.min(1.0, Math.pow(peakRms, 0.28)) : 0
+        resolve({
+          averageLevel: avgLevel,
+          peakLevel: peakLevel,
+          hasAudio: avgRms > 0.001,
+        })
+      })
+    })
+  }
+
   /**
    * 수신된 오디오 청크를 60ms 프레임(1920 bytes) 단위로 분할하여 emit한다.
    * 잔여 바이트는 다음 청크와 결합한다.