diff --git a/src/sounddiff/formats.py b/src/sounddiff/formats.py
index 15564a5..2e03462 100644
--- a/src/sounddiff/formats.py
+++ b/src/sounddiff/formats.py
@@ -21,6 +21,15 @@
 # Formats that require ffmpeg
 FFMPEG_FORMATS = {".mp3", ".aac", ".m4a", ".wma", ".opus"}
 
+# Mapping for original format display names in metadata
+FORMAT_DISPLAY_NAMES: dict[str, str] = {
+    ".mp3": "MP3",
+    ".aac": "AAC",
+    ".m4a": "AAC",
+    ".wma": "WMA",
+    ".opus": "Opus",
+}
+
 
 def _remove_if_exists(path: str) -> None:
     """Remove a file if it exists, silently ignoring missing files."""
@@ -100,6 +109,12 @@ def load_audio(path: str | Path) -> tuple[np.ndarray, AudioMetadata]:
     except RuntimeError as e:
         raise RuntimeError(f"Cannot read audio file: {original_filepath} ({e})") from e
 
+    if original_filepath != read_filepath:
+        ext = original_filepath.suffix.lower()
+        display_format = FORMAT_DISPLAY_NAMES.get(ext, ext.lstrip(".").upper())
+    else:
+        display_format = info.format
+
     try:
         file_size: int | None = original_filepath.stat().st_size
     except OSError:
@@ -111,7 +126,7 @@ def load_audio(path: str | Path) -> tuple[np.ndarray, AudioMetadata]:
         sample_rate=sample_rate,
         channels=data.shape[1],
         bit_depth=_subtype_to_bits(info.subtype),
-        format_name=info.format,
+        format_name=display_format,
         frames=len(data),
         file_size=file_size,
     )
diff --git a/tests/test_formats.py b/tests/test_formats.py
index 89bc58c..f87fb84 100644
--- a/tests/test_formats.py
+++ b/tests/test_formats.py
@@ -56,6 +56,35 @@ def test_frames_matches_data_length(self) -> None:
         data, meta = load_audio(FIXTURES_DIR / "sine_a.wav")
         assert meta.frames == len(data)
 
+    def test_transcoded_format_name_is_correct(self, tmp_path: Path) -> None:
+        """Verifies that transcoded files show their original format name, not WAV."""
+        from unittest.mock import patch
+
+        # 1. Creamos un archivo MP3 falso para pasar la validación de path.exists()
+        fake_mp3 = tmp_path / "test.mp3"
+        fake_mp3.write_text("fake audio content")
+
+        # 2. Simulamos que ffmpeg existe y que la lectura del WAV temporal funciona
+        with (
+            patch("sounddiff.formats.shutil.which", return_value="ffmpeg"),
+            patch("sounddiff.formats.subprocess.run"),
+            patch("sounddiff.formats.sf.info") as mock_info,
+            patch("sounddiff.formats.sf.read") as mock_read,
+        ):
+            # Configuramos el mock para simular lo que devolvería el WAV temporal
+            class MockInfo:
+                format = "WAV"
+                subtype = "PCM_16"
+
+            mock_info.return_value = MockInfo()
+            mock_read.return_value = (np.zeros((100, 2), dtype=np.float64), 44100)
+
+            # 3. Llamamos a tu función
+            _, meta = load_audio(fake_mp3)
+
+            # 4. LA COMPROBACIÓN FINAL: Debe decir MP3 y no WAV
+            assert meta.format_name == "MP3"
+
     def test_file_size_is_populated(self) -> None:
         _, meta = load_audio(FIXTURES_DIR / "sine_a.wav")
         assert meta.file_size is not None