diff --git a/README.md b/README.md
index 025920b0..3f516db9 100644
--- a/README.md
+++ b/README.md
@@ -28,7 +28,7 @@ Instead, herein is proposed ULID:
 - 1.21e+24 unique ULIDs per millisecond
 - Lexicographically sortable!
 - Canonically encoded as a 26 character string, as opposed to the 36 character UUID
-- Uses [Crockford's base32](https://www.crockford.com/base32.html) for better efficiency and readability (5 bits per character) # See also exists issues in [Note](#note)
+- Uses [Crockford's base32](https://www.crockford.com/base32.html) for better efficiency and readability (5 bits per character)
 - Case insensitive
 - No special characters (URL safe)
 - Monotonic sort order (correctly detects and handles the same millisecond)
@@ -326,6 +326,34 @@ ULID.sample(10, period: ulid1.to_time..ulid2.to_time)
 #  ULID(2021-04-28 15:05:06.808 UTC: 01F4CG68ZRST94T056KRZ5K9S4)]
 ```
 
+### ULID specification ambiguity around orthographical variants of the format
+
+I'm afraid so, we should consider [Current ULID spec](https://github.com/ulid/spec/tree/d0c7170df4517939e70129b4d6462cc162f2d5bf#universally-unique-lexicographically-sortable-identifier) has `orthographical variants of the format` possibilities.
+
+>Uses Crockford's base32 for better efficiency and readability (5 bits per character)
+
+The original `Crockford's base32` maps `I`, `L` to `1`, `O` to `0`.
+And accepts freestyle inserting `Hyphens (-)`.
+To consider this patterns or not is different in each implementations.
+
+Current parser/validator/matcher aims to cover `subset of Crockford's base32`.
+I have suggested it would be clarified in [ulid/spec#57](https://github.com/ulid/spec/pull/57).
+
+>Case insensitive
+
+I can understand it might be considered in actual use-case.
+But it is a controversial point, discussing in [ulid/spec#3](https://github.com/ulid/spec/issues/3).
+
+Be that as it may, this gem provides API for handling the nasty possibilities.
+
+`ULID.normalize` and `ULID.normalized?`
+
+```ruby
+ULID.normalize('-olarz3-noekisv4rrff-q6ig5fav--') #=> "01ARZ3N0EK1SV4RRFFQ61G5FAV"
+ULID.normalized?('-olarz3-noekisv4rrff-q6ig5fav--') #=> false
+ULID.normalized?('01ARZ3N0EK1SV4RRFFQ61G5FAV') #=> true
+```
+
 ### UUIDv4 converter for migration use-cases
 
 `ULID.from_uuidv4` and `ULID#to_uuidv4` is the converter.
@@ -418,4 +446,3 @@ The results are not something to be proud of.
 ## Note
 
 - Another choices for sortable and randomness IDs, [UUIDv6, UUIDv7, UUIDv8 might be the one. (But they are still in draft state)](https://www.ietf.org/archive/id/draft-peabody-dispatch-new-uuid-format-01.html), I will track them in [ruby-ulid#37](https://github.com/kachick/ruby-ulid/issues/37)
-- Current parser/validator/matcher aims to cover `subset of Crockford's base32`. Suggesting it in [ulid/spec#57](https://github.com/ulid/spec/pull/57). Be that as it may, I might provide special handler or converter for the exception in [ruby-ulid#57](https://github.com/kachick/ruby-ulid/issues/57) and/or [ruby-ulid#78](https://github.com/kachick/ruby-ulid/issues/78)
diff --git a/lib/ulid.rb b/lib/ulid.rb
index 552e2fa4..4bd5a47c 100644
--- a/lib/ulid.rb
+++ b/lib/ulid.rb
@@ -259,6 +259,27 @@ def self.parse(string)
     from_integer(CrockfordBase32.decode(string))
   end
 
+  # @param [String, #to_str] string
+  # @return [String]
+  # @raise [ParserError] if the given format is not correct for ULID specs, even if ignored `orthographical variants of the format`
+  def self.normalize(string)
+    string = String.try_convert(string)
+    raise ArgumentError, 'ULID.normalize takes only strings' unless string
+
+    normalized_in_crockford = CrockfordBase32.normalize(string)
+    # Ensure the ULID correctness, because CrockfordBase32 does not always mean to satisfy ULID format
+    parse(normalized_in_crockford).to_s
+  end
+
+  # @return [Boolean]
+  def self.normalized?(object)
+    normalized = normalize(object)
+  rescue Exception
+    false
+  else
+    normalized == object
+  end
+
   # @return [Boolean]
   def self.valid?(object)
     string = String.try_convert(object)
diff --git a/lib/ulid/crockford_base32.rb b/lib/ulid/crockford_base32.rb
index 65f2a053..6057b89c 100644
--- a/lib/ulid/crockford_base32.rb
+++ b/lib/ulid/crockford_base32.rb
@@ -51,6 +51,14 @@ class SetupError < ScriptError; end
     CROCKFORD_BASE32_CHAR_BY_N32_CHAR = N32_CHAR_BY_CROCKFORD_BASE32_CHAR.invert.freeze
     N32_CHAR_PATTERN = /[#{CROCKFORD_BASE32_CHAR_BY_N32_CHAR.keys.join}]/.freeze
 
+    VARIANT_BY_STANDARD = {
+      'L' => '1',
+      'I' => '1',
+      'O' => '0',
+      '-' => ''
+    }.freeze
+    VARIANT_PATTERN = /[#{VARIANT_BY_STANDARD.keys.join}]/.freeze
+
     # @api private
     # @param [String] string
     # @return [Integer]
@@ -66,5 +74,12 @@ def self.encode(integer)
       n32encoded = integer.to_s(32)
       n32encoded.upcase.gsub(N32_CHAR_PATTERN, CROCKFORD_BASE32_CHAR_BY_N32_CHAR).rjust(ENCODED_LENGTH, '0')
     end
+
+    # @api private
+    # @param [String] string
+    # @return [String]
+    def self.normalize(string)
+      string.upcase.gsub(VARIANT_PATTERN, VARIANT_BY_STANDARD)
+    end
   end
 end
diff --git a/sig/ulid.rbs b/sig/ulid.rbs
index a933b007..9ff8a688 100644
--- a/sig/ulid.rbs
+++ b/sig/ulid.rbs
@@ -43,12 +43,17 @@ class ULID < Object
     CROCKFORD_BASE32_CHAR_PATTERN: Regexp
     CROCKFORD_BASE32_CHAR_BY_N32_CHAR: Hash[String, String]
     N32_CHAR_PATTERN: Regexp
+    VARIANT_BY_STANDARD: Hash[String, String]
+    VARIANT_PATTERN: Regexp
 
     # A pribate API. Should not be used in your code.
     def self.encode: (Integer integer) -> String
 
     # A pribate API. Should not be used in your code.
     def self.decode: (String string) -> Integer
+
+    # A pribate API. Should not be used in your code.
+    def self.normalize: (String string) -> String
   end
 
   class MonotonicGenerator
@@ -317,6 +322,28 @@ class ULID < Object
                  | (Integer number, ?period: period) -> Array[self]
   def self.valid?: (untyped) -> bool
 
+  # Returns normalized string
+  #
+  # ```ruby
+  # ULID.normalize('-olarz3-noekisv4rrff-q6ig5fav--') #=> "01ARZ3N0EK1SV4RRFFQ61G5FAV"
+  # ULID.normalized?('-olarz3-noekisv4rrff-q6ig5fav--') #=> false
+  # ULID.normalized?('01ARZ3N0EK1SV4RRFFQ61G5FAV') #=> true
+  # ```
+  #
+  # See also [ulid/spec#57](https://github.com/ulid/spec/pull/57) and [ulid/spec#3](https://github.com/ulid/spec/issues/3)
+  def self.normalize: (_ToStr string) -> String
+
+  # Returns `true` if it is normalized string
+  #
+  # ```ruby
+  # ULID.normalize('-olarz3-noekisv4rrff-q6ig5fav--') #=> "01ARZ3N0EK1SV4RRFFQ61G5FAV"
+  # ULID.normalized?('-olarz3-noekisv4rrff-q6ig5fav--') #=> false
+  # ULID.normalized?('01ARZ3N0EK1SV4RRFFQ61G5FAV') #=> true
+  # ```
+  #
+  # See also [ulid/spec#57](https://github.com/ulid/spec/pull/57) and [ulid/spec#3](https://github.com/ulid/spec/issues/3)
+  def self.normalized?: (untyped) -> bool
+
   # Returns parsed ULIDs from given String for rough operations.
   #
   # ```ruby
diff --git a/test/core/test_ulid_class.rb b/test/core/test_ulid_class.rb
index 81cbcc79..0a8ed399 100644
--- a/test/core/test_ulid_class.rb
+++ b/test/core/test_ulid_class.rb
@@ -114,6 +114,98 @@ def test_valid?
     end
   end
 
+  def test_normalize
+    # This is the core of this feature
+    assert_equal(ULID.parse('01ARZ3N0EK1SV4RRFFQ61G5FAV'), ULID.parse(ULID.normalize('-OlARZ3-NoEKISV4rRFF-Q6iG5FAV--')))
+    assert_equal(ULID.parse('01ARZ3N0EK1SV4RRFFQ61G5FAV').to_s, ULID.normalize('-olarz3-noekisv4rrff-q6ig5fav--'))
+
+    normalized = '01ARZ3NDEKTSV4RRFFQ69G5FAV'
+    downcased = normalized.downcase
+    dup_downcased = downcased.dup
+
+    assert(normalized.frozen?)
+    assert_not_same(normalized, ULID.normalize(normalized))
+
+    # This behavior is controversial, should be return non frozen string?
+    assert do
+      ULID.normalize(normalized).frozen?
+    end
+
+    # Ensure the string is not modified in parser
+    assert_equal(false, downcased.frozen?)
+    assert_not_same(downcased, ULID.normalize(downcased))
+    assert_equal(dup_downcased, downcased)
+
+    assert_equal(normalized, ULID.normalize(downcased))
+    assert_instance_of(String, ULID.normalize(downcased))
+
+    # This encoding handling is controversial, should be return original encoding?
+    assert_equal(Encoding::UTF_8, downcased.encoding)
+    assert_equal(Encoding::US_ASCII, ULID.normalize(downcased).encoding)
+
+    [
+      '',
+      "01ARZ3NDEKTSV4RRFFQ69G5FAV\n",
+      '01ARZ3NDEKTSV4RRFFQ69G5FAU',
+      '01ARZ3NDEKTSV4RRFFQ69G5FA',
+      '80000000000000000000000000'
+    ].each do |invalid|
+      err = assert_raises(ULID::ParserError) do
+        ULID.normalize(invalid)
+      end
+      assert_match(/does not match to/, err.message)
+    end
+
+    ULID.sample(1000).each do |sample|
+      assert_equal(sample.to_s, ULID.normalize(sample.to_s))
+      assert_equal(sample.to_s, ULID.normalize(sample.to_s.downcase))
+    end
+
+    assert_raises(ArgumentError) do
+      ULID.normalize
+    end
+
+    [nil, 42, normalized.to_sym, BasicObject.new, Object.new, ULID.parse(normalized)].each do |evil|
+      err = assert_raises(ArgumentError) do
+        ULID.normalize(evil)
+      end
+      assert_equal('ULID.normalize takes only strings', err.message)
+    end
+  end
+
+  def test_normalized?
+    nasty = '-olarz3-noekisv4rrff-q6ig5fav--'
+    assert_equal(false, ULID.normalized?(nasty))
+    assert_equal(true, ULID.normalized?(ULID.normalize(nasty)))
+
+    normalized = '01ARZ3NDEKTSV4RRFFQ69G5FAV'
+    assert_equal(true, ULID.normalized?(normalized))
+    assert_equal(false, ULID.normalized?(normalized.downcase))
+
+    [
+      '',
+      "01ARZ3NDEKTSV4RRFFQ69G5FAV\n",
+      '01ARZ3NDEKTSV4RRFFQ69G5FAU',
+      '01ARZ3NDEKTSV4RRFFQ69G5FA',
+      '80000000000000000000000000'
+    ].each do |invalid|
+      assert_equal(false, ULID.normalized?(invalid))
+    end
+
+    ULID.sample(1000).each do |sample|
+      assert_equal(true, ULID.normalized?(sample.to_s))
+      assert_equal(false, ULID.normalized?(sample.to_s.downcase))
+    end
+
+    assert_raises(ArgumentError) do
+      ULID.normalized?
+    end
+
+    [nil, 42, normalized.to_sym, BasicObject.new, Object.new, ULID.parse(normalized)].each do |evil|
+      assert_equal(false, ULID.normalized?(evil))
+    end
+  end
+
   def test_range
     time_has_more_value_than_milliseconds1 = Time.at(946684800, Rational('123456.789')) # 2000-01-01 00:00:00.123456789 UTC
     time_has_more_value_than_milliseconds2 = Time.at(1620045632, Rational('123456.789')) # 2021-05-03 12:40:32.123456789 UTC