diff --git a/.tool-versions b/.tool-versions
new file mode 100644
index 0000000..b434626
--- /dev/null
+++ b/.tool-versions
@@ -0,0 +1 @@
+gleam 0.32.4
diff --git a/src/glubs/webvtt.gleam b/src/glubs/webvtt.gleam
index 2c61ec6..61d9eef 100644
--- a/src/glubs/webvtt.gleam
+++ b/src/glubs/webvtt.gleam
@@ -4,6 +4,100 @@ import gleam/result
 import gleam/list
 import gleam/int
 
+pub type Cue {
+  Cue(
+    id: option.Option(String),
+    start_time: Int,
+    end_time: Int,
+    payload: String,
+  )
+}
+
+pub type WebVTT {
+  WebVTT(comment: option.Option(String), cues: List(Cue))
+}
+
+pub type ParserError {
+  ParserError(String)
+}
+
+pub fn parse(webvtt: String) -> Result(WebVTT, ParserError) {
+  let [header, ..cues] =
+    webvtt
+    |> string.replace("\r\n", "\n")
+    |> string.split("\n\n")
+
+  // TODO: Metadata still needs to be parsed
+  let [header, ..] = string.split(header, "\n")
+
+  use comment <- result.try(parse_comment(header))
+  use cues <- result.try(list.try_map(cues, parse_cue))
+
+  Ok(WebVTT(comment: comment, cues: cues))
+}
+
+pub fn to_string(_webvtt: WebVTT) -> String {
+  todo
+}
+
+fn parse_comment(header: String) -> Result(option.Option(String), ParserError) {
+  case header {
+    "WEBVTT" -> Ok(None)
+    "WEBVTT\t" <> comment -> Ok(Some(comment))
+    "WEBVTT " <> comment -> Ok(Some(comment))
+    "WEBVTT" <> _other ->
+      Error(ParserError("Header comment must start with space or tab"))
+    _other -> Error(ParserError("Must start with \"WEBVTT\""))
+  }
+}
+
+fn parse_cue(cue: String) -> Result(Cue, ParserError) {
+  use #(id, rest) <- result.try(parse_cue_id(cue))
+
+  case string.split_once(rest, "\n") {
+    Ok(#(line, payload)) -> {
+      use #(start, end) <- result.try(parse_timestamps(line))
+      Ok(Cue(id: id, payload: payload, start_time: start, end_time: end))
+    }
+    Error(Nil) -> Error(ParserError("Invalid cue"))
+  }
+}
+
+fn parse_cue_id(
+  cue: String,
+) -> Result(#(option.Option(String), String), ParserError) {
+  case string.split_once(cue, "\n") {
+    Ok(#(id, rest)) -> {
+      case string.contains(id, "-->") {
+        True -> Ok(#(None, cue))
+        False -> Ok(#(Some(id), rest))
+      }
+    }
+    Error(Nil) -> Error(ParserError("Invalid cue"))
+  }
+}
+
+fn parse_timestamps(line: String) -> Result(#(Int, Int), ParserError) {
+  case string.split(line, " --> ") {
+    [start, end] -> {
+      use start <- result.try(
+        start
+        |> parse_timestamp()
+        |> result.replace_error(ParserError("Invalid start timestamp")),
+      )
+
+      use end <- result.try(
+        end
+        |> parse_timestamp()
+        |> result.replace_error(ParserError("Invalid end timestamp")),
+      )
+
+      Ok(#(start, end))
+    }
+    _other -> Error(ParserError("Invalid timestamp"))
+  }
+}
+
 pub type Token {
   StartTag(
     tag: String,
diff --git a/test/glubs/webvtt_test.gleam b/test/glubs/webvtt_test.gleam
index df404f6..a387302 100644
--- a/test/glubs/webvtt_test.gleam
+++ b/test/glubs/webvtt_test.gleam
@@ -1,6 +1,41 @@
 import gleeunit/should
 import gleam/option.{None, Some}
-import glubs/webvtt.{EndTag, StartTag, Text, Timestamp}
+import glubs/webvtt.{Cue, EndTag, StartTag, Text, Timestamp, WebVTT}
+
+pub fn parse_invalid_header_test() {
+  "INVALID"
+  |> webvtt.parse()
+  |> should.equal(Error(webvtt.ParserError("Must start with \"WEBVTT\"")))
+}
+
+pub fn parse_attached_header_test() {
+  "WEBVTTinvalid"
+  |> webvtt.parse()
+  |> should.equal(Error(webvtt.ParserError(
+    "Header comment must start with space or tab",
+  )))
+}
+
+pub fn parse_only_header_test() {
+  "WEBVTT"
+  |> webvtt.parse()
+  |> should.be_ok()
+}
+
+pub fn parse_header_with_comment_test() {
+  "WEBVTT This is a comment"
+  |> webvtt.parse()
+  |> should.equal(Ok(WebVTT(comment: Some("This is a comment"), cues: [])))
+}
+
+pub fn parse_cue_test() {
+  "WEBVTT\n\n1\n00:00.123 --> 00:00.456\nTest"
+  |> webvtt.parse()
+  |> should.equal(Ok(WebVTT(
+    comment: None,
+    cues: [Cue(id: Some("1"), start_time: 123, end_time: 456, payload: "Test")],
+  )))
+}
 
 pub fn tokenize_text_test() {
   "Hello"