diff --git a/README.md b/README.md
index 792e33e..84c89bb 100644
--- a/README.md
+++ b/README.md
@@ -6,12 +6,105 @@ Generates and compiles mapping files for Via*. Current mapping files can be foun
 
 Compile the project using `mvn clean build` and put the jar in some directory, ideally the project root.
 
-Then run the jar with: `java -jar MappingsGenerator.jar <path to server jar> <version>`,
-e.g. `java -jar MappingsGenerator.jar server.jar 20w22a`. The mapping file will then be generated in the `mappings/`
-directory.
+Then run the jar with:
+
+```
+java -jar MappingsGenerator.jar <path to server jar> <version>
+```
+
+The mapping file will then be generated in the `mappings/` directory.
+
+### Directly compiling them to compact files
+
+Optionally, you can pass a third argument to have it generate the compact nbt mappings file as
+well:
+
+```
+java -jar MappingsGenerator.jar server.jar <version> <from version>
+```
+
+e.g. to generate the compact nbt mappings for 1.19.3->1.19.4:
+
+```
+java -jar MappingsGenerator.jar server.jar 1.19.4 1.19.3
+```
 
 ## Compiling json mapping files into compact nbt files
 
+If you want to generate the compact mapping files with already present json files, you can also trigger the optimizer on
+its own by starting the `MappingsOptimizer` class with the two arguments flipped:
+
+```
+java -DbundlerMainClass=com.viaversion.mappinggenerator.MappingsOptimizer -jar <from version> <to version>
+```
+
+## Json format
+
+The json files contain a number of Minecraft registries in form of json arrays, where the index corresponds to the id of
+the entry.
+
+Diff files for either ViaVersion or ViaBackwards then contain additional entries for changed identifiers, either in form
+of string→string or int→string mappings. These files need to be manually filled. If any such entries are required, the
+optimizer will give a warning with the missing keys.
+
+## Compact format
+
+Compact files are always saved as [NBT](https://minecraft.fandom.com/wiki/NBT_format). ViaVersion uses its
+own [OpenNBT](https://github.com/ViaVersion/OpenNBT) as its NBT reader/writer.
+
+### Identifier files
+
+Next to a standardized compact format, there are extra files per version to store the identifier arrays of certain
+registries, as their full keys might be required. These files simply contain any number of json arrays.
+
+### Mapping files
+
+Each mapping file contains a `v` int tag with the format version, currently being `1`.
+
+In each mapping file, a number of extra objects may be contained, such as string→string mappings for sounds. Most other
+parts (including blockstates, blocks, items, blockentities, enchantments, paintings, entities, particles, argumenttypes,
+and statistics) are stored as compound tags, containing:
+
+* `id` (byte tag) determining the storage type as defined below
+* `size` (int tag) the number of unmapped entries in the registry
+* `mappedSize` (int tag) the number of mapped entries in the registry
+
+The rest of the content depends on the storage type, each resulting in vastly different storage sizes depending on the
+number and distribution of id changes, used to make the mapping files about as small as possible.
+
+### Direct value storage
+
+The direct storage simply stores an array of ints exactly as they can be used in the protocol.
+
+* `id` (byte tag) is `0`
+* `val` (int array tag) contains the mapped ids, where their array index corresponds to the unmapped id
+
+### Changed value storage
+
+The changed value storage stores two int arrays: One containing the changed unmapped ids, and one their corresponding
+mapped ids in a simple int→int mapping over the two arrays.
+
+* `id` (byte tag) is `1`
+* `at` (int array tag) contains the unmapped ids that have been changed
+* `val` (int array tag) contains the mapped ids, indexed by the same index as the unmapped id in `at`
+* Optional: `nofill` (byte tag): Unless present, all ids between the ones found in `at` are mapped to their identity
+
+### Shifted value storage
+
+The shifted value storage stores two int arrays: One containing the unmapped ids that end a sequence of mapped ids. For
+an index `i`, all unmapped ids between `at[i] + sequence` (inclusive) and `at[i + 1]` (exclusive) are mapped to `to[i] + sequence`.
+
+* `id` (byte tag) is `2`
+* `at` (int array tag) contains the unmapped ids, where their mapped is is *not* simply the last mapped id + 1
+* `to` (int array tag) contains the mapped ids, indexed by the same index as the unmapped id in `at`
+
+### Identity storage
+
+The identity storage signifies that every id between `0` and `size` is mapped to itself. This is sometimes used over
+simply leaving out the entry to make sure ids stay in bounds.
+
+* `id` (byte tag) is `3`
+
 ## License
 
 The Java code is licensed under the GNU GPL v3 license. The files under `mappings/` are free to copy, use, and expand
diff --git a/src/main/java/com/viaversion/mappingsgenerator/MappingsGenerator.java b/src/main/java/com/viaversion/mappingsgenerator/MappingsGenerator.java
index d1dcb21..9f59bda 100644
--- a/src/main/java/com/viaversion/mappingsgenerator/MappingsGenerator.java
+++ b/src/main/java/com/viaversion/mappingsgenerator/MappingsGenerator.java
@@ -77,7 +77,7 @@ public final class MappingsGenerator {
         MappingsGenerator.collectMappings(version);
 
         if (versionToCompareFrom != null) {
-            LOGGER.info("Running mappings optimizer for versions {} -> {}...", versionToCompareFrom, version);
+            LOGGER.info("Running mappings optimizer for versions {} → {}...", versionToCompareFrom, version);
             MappingsOptimizer.optimizeAndSaveAsNBT(versionToCompareFrom, version);
         }
     }
diff --git a/src/main/java/com/viaversion/mappingsgenerator/MappingsOptimizer.java b/src/main/java/com/viaversion/mappingsgenerator/MappingsOptimizer.java
index 3541c24..b849ba2 100644
--- a/src/main/java/com/viaversion/mappingsgenerator/MappingsOptimizer.java
+++ b/src/main/java/com/viaversion/mappingsgenerator/MappingsOptimizer.java
@@ -89,7 +89,7 @@ public final class MappingsOptimizer {
      * @param to   version to map to
      */
     public static void optimizeAndSaveAsNBT(final String from, final String to) throws IOException {
-        LOGGER.info("Compacting json mapping files for versions {} -> {}...", to, from);
+        LOGGER.info("Compacting json mapping files for versions {} → {}...", to, from);
         final JsonObject unmappedObject = MappingsLoader.load(MAPPING_FILE_FORMAT.formatted(from));
         final JsonObject mappedObject = MappingsLoader.load(MAPPING_FILE_FORMAT.formatted(to));
         final JsonObject diffObject = MappingsLoader.load(DIFF_FILE_FORMAT.formatted(from, to));