diff --git a/internal/ingestion/component/media_dispatch_test.go b/internal/ingestion/component/media_dispatch_test.go
index 2c465616cc..ee98c2bf73 100644
--- a/internal/ingestion/component/media_dispatch_test.go
+++ b/internal/ingestion/component/media_dispatch_test.go
@@ -25,6 +25,7 @@ import (
 	"ragflow/internal/common"
 	"ragflow/internal/entity"
 	modelModule "ragflow/internal/entity/models"
+	"ragflow/internal/ingestion/component/schema"
 	"ragflow/internal/utility"
 
 	"gorm.io/gorm"
@@ -362,9 +363,13 @@ func TestMaybeDispatchAudio_TextCarriesTranscription(t *testing.T) {
 	}
 }
 
-// TestMaybeDispatchAudio_DefaultOutputFormatJson covers Parser 2.11:
-// the default audio output_format must be "json" (matching Python
-// parser.py:232 and AllowedOutputFormat["audio"]={"json"}).
+// TestMaybeDispatchAudio_DefaultOutputFormatJson covers the
+// maybeDispatchAudio fallback: when an audio setup omits
+// output_format entirely, the dispatch defaults to "json" and wraps
+// the transcription as a JSON item. (The defaultSetups value is
+// "text" to mirror the Python audio setup in parser.py; this test
+// deliberately supplies an empty setup to exercise the fallback
+// inside the dispatch itself.)
 func TestMaybeDispatchAudio_DefaultOutputFormatJson(t *testing.T) {
 	const want = "hello world"
 	drv := &audioTranscribeDriver{transcription: want}
@@ -373,8 +378,9 @@ func TestMaybeDispatchAudio_DefaultOutputFormatJson(t *testing.T) {
 	resolveTenantModelByType = func(ctx context.Context, db *gorm.DB, tenantID string, modelType entity.ModelType) (modelModule.ModelDriver, string, *modelModule.APIConfig, int, error) {
 		return drv, "asr-model", &modelModule.APIConfig{}, 0, nil
 	}
-	setups := defaultSetups()
-	// Do NOT set output_format — exercise the default path.
+	// No output_format key — exercise the default path inside
+	// maybeDispatchAudio.
+	setups := map[string]schema.ParserSetup{"audio": {}}
 	res, dispatched, err := maybeDispatchAudio(
 		context.Background(),
 		nil,
diff --git a/internal/ingestion/component/parser.go b/internal/ingestion/component/parser.go
index 5031a5b4be..19c74443f0 100644
--- a/internal/ingestion/component/parser.go
+++ b/internal/ingestion/component/parser.go
@@ -314,7 +314,7 @@ func defaultSetups() map[string]schema.ParserSetup {
 				"aiff", "au", "midi", "wma", "realaudio", "vqf",
 				"oggvorbis", "ape",
 			},
-			"output_format": "json",
+			"output_format": "text",
 		},
 		"video": {
 			"suffix":        []string{"mp4", "avi", "mkv"},
diff --git a/internal/ingestion/component/parser_dispatch_test.go b/internal/ingestion/component/parser_dispatch_test.go
index 35036258e4..f1f9badb43 100644
--- a/internal/ingestion/component/parser_dispatch_test.go
+++ b/internal/ingestion/component/parser_dispatch_test.go
@@ -305,6 +305,84 @@ func TestDefaultSetups_DOCX_OutputFormatMarkdown(t *testing.T) {
 	}
 }
 
+// TestResolveOutputFormat_AudioOutputFormats pins the audio-family
+// whitelist against the builtin audio template. The template
+// (ingestion_pipeline_audio.json) and the Python default audio setup
+// (rag/flow/parser/parser.py) both use output_format="text" — an audio
+// transcription is inherently plain text. "text" must therefore pass
+// the whitelist, "json" must stay accepted, and a format outside the
+// whitelist must still be rejected.
+func TestResolveOutputFormat_AudioOutputFormats(t *testing.T) {
+	allowed := schema.ParserParam{}.Defaults().AllowedOutputFormat
+	audioAllowed, ok := allowed["audio"]
+	if !ok {
+		t.Fatal("allowed_output_format: audio key missing")
+	}
+	has := func(want string) bool {
+		for _, v := range audioAllowed {
+			if strings.EqualFold(v, want) {
+				return true
+			}
+		}
+		return false
+	}
+	if !has("text") {
+		t.Errorf("allowed_output_format[audio] = %v, want it to include %q (builtin audio template and Python default use it)", audioAllowed, "text")
+	}
+	if !has("json") {
+		t.Errorf("allowed_output_format[audio] = %v, want it to include %q", audioAllowed, "json")
+	}
+
+	cases := []struct {
+		name    string
+		format  string
+		wantErr bool
+	}{
+		{name: "text accepted", format: "text"},
+		{name: "json accepted", format: "json"},
+		{name: "html rejected", format: "html", wantErr: true},
+	}
+	for _, tc := range cases {
+		t.Run(tc.name, func(t *testing.T) {
+			setups := map[string]schema.ParserSetup{"audio": {"output_format": tc.format}}
+			got, err := resolveOutputFormat("audio", setups, allowed)
+			if tc.wantErr {
+				if err == nil {
+					t.Fatalf("want error for audio output_format=%q, got %q", tc.format, got)
+				}
+				return
+			}
+			if err != nil {
+				t.Fatalf("audio output_format=%q: unexpected error: %v", tc.format, err)
+			}
+			if got != tc.format {
+				t.Errorf("got %q, want %q", got, tc.format)
+			}
+		})
+	}
+}
+
+// TestDefaultSetups_Audio_OutputFormatText pins the audio default to
+// "text", matching the Python default setup
+// (rag/flow/parser/parser.py audio block) and the builtin audio
+// template. The default feeds both the whitelist gate and the ASR
+// dispatch, so it must not drift to a format the audio pipeline does
+// not produce.
+func TestDefaultSetups_Audio_OutputFormatText(t *testing.T) {
+	setups := defaultSetups()
+	audio, ok := setups["audio"]
+	if !ok {
+		t.Fatal("defaultSetups: audio key missing")
+	}
+	got, ok := audio["output_format"].(string)
+	if !ok {
+		t.Fatal("defaultSetups: audio.output_format missing or not a string")
+	}
+	if got != "text" {
+		t.Errorf("audio.output_format = %q, want %q", got, "text")
+	}
+}
+
 func TestConfigureParserFromSetups_UsesPythonFamilySetup(t *testing.T) {
 	setups := defaultSetups()
 	got := &captureSetupConfigurer{}
diff --git a/internal/ingestion/component/schema/parser.go b/internal/ingestion/component/schema/parser.go
index d47ed57df9..5f6e91d1fc 100644
--- a/internal/ingestion/component/schema/parser.go
+++ b/internal/ingestion/component/schema/parser.go
@@ -85,7 +85,7 @@ func (ParserParam) Defaults() ParserParam {
 			"markdown":    {"text", "json"},
 			"text&code":   {"text", "json"},
 			"html":        {"text", "json"},
-			"audio":       {"json"},
+			"audio":       {"text", "json"},
 			"video":       {},
 			"epub":        {"text", "json"},
 			"json":        {"json"},