diff --git a/scripts/autogen.sh b/scripts/autogen.sh
new file mode 100644
index 0000000..602a45d
--- /dev/null
+++ b/scripts/autogen.sh
@@ -0,0 +1,15 @@
+#!/bin/sh
+
+# Download libs
+apk add --no-cache curl jq
+npm install --save-dev xml-js
+
+# Download spec from source of truth
+curl https://raw.githubusercontent.com/ietf-wg-cellar/matroska-specification/master/ebml_matroska.xml -o ebml_matroska.xml
+
+# Generate normalised schema
+node convert.js | jq . > converted.json
+node baseline.js | jq . > baselined.json
+
+# Compare
+#diff --color=always -u baselined.json converted.json | less -r
diff --git a/scripts/baseline.js b/scripts/baseline.js
new file mode 100644
index 0000000..d8f98cf
--- /dev/null
+++ b/scripts/baseline.js
@@ -0,0 +1,3 @@
+const schema = require("../lib/schema.js")
+
+console.log(JSON.stringify(schema.byEbmlID))
diff --git a/scripts/convert.js b/scripts/convert.js
new file mode 100644
index 0000000..473fb37
--- /dev/null
+++ b/scripts/convert.js
@@ -0,0 +1,85 @@
+const convert = require("xml-js"),
+  fs = require("fs")
+
+const mapType = type => {
+  switch(type) {
+    case "binary":
+      return "b";
+    case "date":
+      return "d";
+    case "float":
+      return "f";
+    case "integer":
+      return "i";
+    case "master":
+      return "m";
+    case "string":
+      return "s";
+    case "uinteger":
+      return "u";
+    case "utf-8":
+      return "8";
+    default:
+      throw `Type ${type} not found`;
+  }
+}
+
+fs.readFile("ebml_matroska.xml", (err, xml) => {
+  if (err) throw err;
+  const json = convert.xml2js(xml, {compact: true})
+
+  var out = {}
+  for (element of json.EBMLSchema.element) {
+    const key = parseInt(element._attributes["id"], 16)
+
+    const docs = "documentation" in element
+      ? (Array.isArray(element.documentation) ? element.documentation : [element.documentation])
+      : []
+    const extensions = "extension" in element
+      ? (Array.isArray(element.extension) ? element.extension : [element.extension])
+      : []
+
+    let value = {}
+    value["name"] = element._attributes["name"]
+    for (ext of extensions) {
+      if (ext._attributes.type == "libmatroska" && ("cppname" in ext._attributes)) {
+        value["cppname"] = ext._attributes.cppname
+      }
+    }
+    value["level"] = element._attributes["path"].split("\\").length - 2
+    value["type"] = mapType(element._attributes["type"])
+    if ("minOccurs" in element._attributes && element._attributes.minOccurs == 1) {
+      value["mandatory"] = true
+    }
+    if (!("maxOccurs" in element._attributes) || element._attributes.maxOccurs > 1) {
+      value["multiple"] = true
+    }
+    for (attr of ["minver", "maxver"]) {
+      if (attr in element._attributes) {
+        value[attr] = parseInt(element._attributes[attr])
+      }
+    }
+    for (ext of extensions) {
+      if (ext._attributes.type == "webmproject.org" && ext._attributes.webm == "1") {
+        value["webm"] = true
+      }
+      if (ext._attributes.type == "divx.com" && ext._attributes.divx == "1") {
+        value["divx"] = true
+      }
+    }
+    for (attr of ["default", "range"]) {
+      if (attr in element._attributes) {
+        value[attr] = element._attributes[attr]
+      }
+    }
+    for (doc of docs) {
+      if (doc._attributes.purpose == "definition") {
+        value["description"] = doc._text.replace("\n", " ")
+      }
+    }
+
+    out[key] = value
+  }
+
+  console.log(JSON.stringify(out))
+})