diff --git a/cmp/compare_test.go b/cmp/compare_test.go
index 88b7d45..975e4be 100644
--- a/cmp/compare_test.go
+++ b/cmp/compare_test.go
@@ -1320,6 +1320,11 @@ using the AllowUnexported option.`, "\n"),
 		x:      "d5c14bdf6bac81c27afc5429500ed750\n25483503b557c606dad4f144d27ae10b\n90bdbcdbb6ea7156068e3dcfb7459244\n978f480a6e3cced51e297fbff9a506b7\n",
 		y:      "Xd5c14bdf6bac81c27afc5429500ed750\nX25483503b557c606dad4f144d27ae10b\nX90bdbcdbb6ea7156068e3dcfb7459244\nX978f480a6e3cced51e297fbff9a506b7\n",
 		reason: "all lines are different, so diffing based on lines is pointless",
+	}, {
+		label:  label + "/JapaneseUTF8",
+		x:      "プライベート ブランド ジャケット",
+		y:      "プライベート ブランド シャツ",
+		reason: "multi-byte UTF-8 characters should be readable in diff output (issue #314)",
 	}, {
 		label:  label + "/StringifiedBytes",
 		x:      struct{ X []byte }{[]byte("hello, world!")},
diff --git a/cmp/report_slices.go b/cmp/report_slices.go
index 23e444f..349fca2 100644
--- a/cmp/report_slices.go
+++ b/cmp/report_slices.go
@@ -99,7 +99,7 @@ func (opts formatOptions) FormatDiffSlice(v *valueNode) textNode {
 	// Auto-detect the type of the data.
 	var sx, sy string
 	var ssx, ssy []string
-	var isString, isMostlyText, isPureLinedText, isBinary bool
+	var isString, isMostlyText, isPureLinedText, isBinary, isValidUTF8 bool
 	switch {
 	case t.Kind() == reflect.String:
 		sx, sy = vx.String(), vy.String()
@@ -133,6 +133,7 @@ func (opts formatOptions) FormatDiffSlice(v *valueNode) textNode {
 		isMostlyText = float64(numValidRunes) > math.Floor(0.90*float64(numTotalRunes))
 		isPureLinedText = isPureText && numLines >= 4 && maxLineLen <= 1024
 		isBinary = !isMostlyText
+		isValidUTF8 = utf8.ValidString(sx) && utf8.ValidString(sy)
 
 		// Avoid diffing by lines if it produces a significantly more complex
 		// edit script than diffing by bytes.
@@ -251,13 +252,32 @@ func (opts formatOptions) FormatDiffSlice(v *valueNode) textNode {
 	// then perform differencing in approximately fixed-sized chunks.
 	// The output is printed as quoted strings.
 	case isMostlyText:
-		list = opts.formatDiffSlice(
-			reflect.ValueOf(sx), reflect.ValueOf(sy), 64, "byte",
-			func(v reflect.Value, d diffMode) textRecord {
-				s := formatString(v.String())
-				return textRecord{Diff: d, Value: textLine(s)}
-			},
-		)
+		if isValidUTF8 {
+			// Rune-based diffing for valid UTF-8 strings
+			rx := []rune(sx)
+			ry := []rune(sy)
+			list = opts.formatDiffSlice(
+				reflect.ValueOf(rx), reflect.ValueOf(ry), 64, "rune",
+				func(v reflect.Value, d diffMode) textRecord {
+					// Convert []rune chunk back to string for display
+					runes := make([]rune, v.Len())
+					for i := 0; i < v.Len(); i++ {
+						runes[i] = rune(v.Index(i).Int())
+					}
+					s := formatString(string(runes))
+					return textRecord{Diff: d, Value: textLine(s)}
+				},
+			)
+		} else {
+			// Byte-based diffing for invalid UTF-8 (original behavior)
+			list = opts.formatDiffSlice(
+				reflect.ValueOf(sx), reflect.ValueOf(sy), 64, "byte",
+				func(v reflect.Value, d diffMode) textRecord {
+					s := formatString(v.String())
+					return textRecord{Diff: d, Value: textLine(s)}
+				},
+			)
+		}
 
 	// If the text appears to be binary data,
 	// then perform differencing in approximately fixed-sized chunks.
diff --git a/cmp/testdata/diffs b/cmp/testdata/diffs
index be77b95..7fda3ef 100644
--- a/cmp/testdata/diffs
+++ b/cmp/testdata/diffs
@@ -515,7 +515,7 @@
 + 		"21 2nd Street",
   		`","city":"New York","state":"NY","postalCode":"10021-3100"},"pho`,
   		`neNumbers":[{"type":"home","number":"212 555-1234"},{"type":"off`,
-  		... // 101 identical bytes
+  		... // 101 identical runes
   	}, ""),
   	BytesB: nil,
   	BytesC: nil,
@@ -1019,7 +1019,7 @@
 <<< TestDiff/Reporter/LargeStringInInterface
   struct{ X any }{
   	X: strings.Join({
-  		... // 485 identical bytes
+  		... // 485 identical runes
   		"s mus. Pellentesque mi lorem, consectetur id porttitor id, solli",
   		"citudin sit amet enim. Duis eu dolor magna. Nunc ut augue turpis",
 - 		".",
@@ -1030,7 +1030,7 @@
 <<< TestDiff/Reporter/LargeBytesInInterface
   struct{ X any }{
   	X: bytes.Join({
-  		... // 485 identical bytes
+  		... // 485 identical runes
   		"s mus. Pellentesque mi lorem, consectetur id porttitor id, solli",
   		"citudin sit amet enim. Duis eu dolor magna. Nunc ut augue turpis",
 - 		".",
@@ -1098,6 +1098,13 @@
   	"978f480a6e3cced51e297fbff9a506b7\n",
   }, "")
 >>> TestDiff/Reporter/AllLinesDiffer
+<<< TestDiff/Reporter/JapaneseUTF8
+  strings.Join({
+  	"プライベート ブランド ",
+- 	"ジャケット",
++ 	"シャツ",
+  }, "")
+>>> TestDiff/Reporter/JapaneseUTF8
 <<< TestDiff/Reporter/StringifiedBytes
   struct{ X []uint8 }{
 - 	X: []uint8("hello, world!"),