diff --git a/lib/fluent/plugin/in_tail.rb b/lib/fluent/plugin/in_tail.rb
index ef920c081d..914617b3e4 100644
--- a/lib/fluent/plugin/in_tail.rb
+++ b/lib/fluent/plugin/in_tail.rb
@@ -64,14 +64,10 @@ def initialize
     config_param :multiline_flush_interval, :time, default: nil
     desc 'Enable the additional watch timer.'
     config_param :enable_watch_timer, :bool, default: true
+    desc 'The encoding after conversion of the input.'
+    config_param :encoding, :string, default: nil
     desc 'The encoding of the input.'
-    config_param :encoding, default: nil do |encoding_name|
-      begin
-        Encoding.find(encoding_name)
-      rescue ArgumentError => e
-        raise ConfigError, e.message
-      end
-    end
+    config_param :from_encoding, :string, default: nil
     desc 'Add the log path being tailed to records. Specify the field name to be used.'
     config_param :path_key, :string, default: nil
 
@@ -92,6 +88,7 @@ def configure(conf)
 
       configure_parser(conf)
       configure_tag
+      configure_encoding
 
       @multiline_mode = conf['format'] =~ /multiline/
       @receive_handler = if @multiline_mode
@@ -117,6 +114,25 @@ def configure_tag
       end
     end
 
+    def configure_encoding
+      unless @encoding
+        if @from_encoding
+          raise ConfigError, "tail: 'from_encoding' parameter must be specified with 'encoding' parameter."
+        end
+      end
+
+      @encoding = parse_encoding_param(@encoding) if @encoding
+      @from_encoding = parse_encoding_param(@from_encoding) if @from_encoding
+    end
+
+    def parse_encoding_param(encoding_name)
+      begin
+        Encoding.find(encoding_name) if encoding_name
+      rescue ArgumentError => e
+        raise ConfigError, e.message
+      end
+    end
+
     def start
       super
 
@@ -251,7 +267,13 @@ def close_watcher_after_rotate_wait(tw)
     def flush_buffer(tw)
       if lb = tw.line_buffer
         lb.chomp!
-        lb.force_encoding(@encoding) if @encoding
+        if @encoding
+          if @from_encoding
+            lb.encode!(@encoding, @from_encoding)
+          else
+            lb.force_encoding(@encoding)
+          end
+        end
         @parser.parse(lb) { |time, record|
           if time && record
             tag = if @tag_prefix || @tag_suffix
@@ -300,7 +322,13 @@ def receive_lines(lines, tail_watcher)
     def convert_line_to_event(line, es, tail_watcher)
       begin
         line.chomp!  # remove \n
-        line.force_encoding(@encoding) if @encoding
+        if @encoding
+          if @from_encoding
+            line.encode!(@encoding, @from_encoding)
+          else
+            line.force_encoding(@encoding)
+          end
+        end
         @parser.parse(line) { |time, record|
           if time && record
             record[@path_key] ||= tail_watcher.path unless @path_key.nil?
diff --git a/test/plugin/test_in_tail.rb b/test/plugin/test_in_tail.rb
index 385ead6f16..872f0fdfde 100644
--- a/test/plugin/test_in_tail.rb
+++ b/test/plugin/test_in_tail.rb
@@ -71,6 +71,32 @@ def test_configure_encoding
     end
   end
 
+  def test_configure_from_encoding
+    # If only specified from_encoding raise ConfigError
+    assert_raise(Fluent::ConfigError) do
+      d = create_driver(SINGLE_LINE_CONFIG + 'from_encoding utf-8')
+    end
+
+    # valid setting
+    d = create_driver %[
+      format /(?<message>.*)/
+      read_from_head true
+      from_encoding utf-8
+      encoding utf-8
+    ]
+    assert_equal Encoding::UTF_8, d.instance.from_encoding
+
+    # invalid from_encoding
+    assert_raise(Fluent::ConfigError) do
+      d = create_driver %[
+        format /(?<message>.*)/
+        read_from_head true
+        from_encoding no-such-encoding
+        encoding utf-8
+      ]
+    end
+  end
+
   # TODO: Should using more better approach instead of sleep wait
 
   def test_emit
@@ -403,6 +429,28 @@ def test_encoding(data)
     assert_equal(encoding, emits[0][2]['message'].encoding)
   end
 
+  def test_from_encoding
+    d = create_driver %[
+      format /(?<message>.*)/
+      read_from_head true
+      from_encoding cp932
+      encoding utf-8
+    ]
+
+    d.run do
+      sleep 1
+
+      File.open("#{TMP_DIR}/tail.txt", "w:cp932") {|f|
+        f.puts "\x82\xCD\x82\xEB\x81\x5B\x82\xED\x81\x5B\x82\xE9\x82\xC7".force_encoding(Encoding::CP932)
+      }
+      sleep 1
+    end
+
+    emits = d.emits
+    assert_equal("\x82\xCD\x82\xEB\x81\x5B\x82\xED\x81\x5B\x82\xE9\x82\xC7".force_encoding(Encoding::CP932).encode(Encoding::UTF_8), emits[0][2]['message'])
+    assert_equal(Encoding::UTF_8, emits[0][2]['message'].encoding)
+  end
+
   # multiline mode test
 
   def test_multiline
@@ -507,6 +555,31 @@ def test_multiline_encoding_of_flushed_record(data)
     end
   end
 
+  def test_multiline_from_encoding_of_flushed_record
+    d = create_driver %[
+      format multiline
+      format1 /^s (?<message1>[^\\n]+)(\\nf (?<message2>[^\\n]+))?(\\nf (?<message3>.*))?/
+      format_firstline /^[s]/
+      multiline_flush_interval 2s
+      read_from_head true
+      from_encoding cp932
+      encoding utf-8
+    ]
+
+    d.run do
+      sleep 1
+      File.open("#{TMP_DIR}/tail.txt", "w:cp932") { |f|
+        f.puts "s \x82\xCD\x82\xEB\x81\x5B\x82\xED\x81\x5B\x82\xE9\x82\xC7".force_encoding(Encoding::CP932)
+      }
+
+      sleep 4
+      emits = d.emits
+      assert_equal(1, emits.length)
+      assert_equal("\x82\xCD\x82\xEB\x81\x5B\x82\xED\x81\x5B\x82\xE9\x82\xC7".force_encoding(Encoding::CP932).encode(Encoding::UTF_8), emits[0][2]['message1'])
+      assert_equal(Encoding::UTF_8, emits[0][2]['message1'].encoding)
+    end
+  end
+
   def test_multiline_with_multiple_formats
     File.open("#{TMP_DIR}/tail.txt", "wb") { |f| }