diff --git a/tensorflow_datasets/core/download/download_manager.py b/tensorflow_datasets/core/download/download_manager.py
index 82362ba068b..675787b04a8 100644
--- a/tensorflow_datasets/core/download/download_manager.py
+++ b/tensorflow_datasets/core/download/download_manager.py
@@ -53,7 +53,7 @@
 ExtractPath = epath.PathLike | resource_lib.Resource
 
 
-def get_downloader(*args: Any, **kwargs: Any):
+def get_downloader(*args: Any, **kwargs: Any) -> downloader._Downloader:
   return downloader.get_downloader(*args, **kwargs)
 
 
diff --git a/tensorflow_datasets/core/download/downloader.py b/tensorflow_datasets/core/download/downloader.py
index 2e4652daec6..ff5fc60e5ed 100644
--- a/tensorflow_datasets/core/download/downloader.py
+++ b/tensorflow_datasets/core/download/downloader.py
@@ -15,6 +15,8 @@
 
 """Async download API with checksum verification. No business logic."""
 
+from __future__ import annotations
+
 from collections.abc import Iterable, Iterator
 import concurrent.futures
 import contextlib
@@ -57,7 +59,7 @@ class DownloadResult:
 
 
 @utils.memoize()
-def get_downloader(*args: Any, **kwargs: Any) -> '_Downloader':
+def get_downloader(*args: Any, **kwargs: Any) -> _Downloader:
   return _Downloader(*args, **kwargs)
 
 
diff --git a/tensorflow_datasets/text/c4.py b/tensorflow_datasets/text/c4.py
index d45187883d8..38d4639495b 100644
--- a/tensorflow_datasets/text/c4.py
+++ b/tensorflow_datasets/text/c4.py
@@ -24,7 +24,6 @@
 
 from absl import logging
 from etils import epath
-from tensorflow_datasets.core.utils.lazy_imports_utils import tensorflow as tf
 import tensorflow_datasets.public_api as tfds
 from tensorflow_datasets.text import c4_utils
 import tree
@@ -499,7 +498,9 @@ def _info(self):
         homepage="https://github.com/google-research/text-to-text-transfer-transformer#datasets",
     )
 
-  def _split_generators(self, dl_manager, pipeline):
+  def _split_generators(
+      self, dl_manager: tfds.download.DownloadManager, pipeline
+  ):
     # We will automatically download the first default CC version, but others
     # need to be manually downloaded.
     cc_versions = set(self.builder_config.cc_versions)
@@ -521,8 +522,8 @@ def _split_generators(self, dl_manager, pipeline):
     file_paths = dl_manager.download_and_extract(files_to_download)
 
     if self.builder_config.webtextlike:
-      owt_path = os.path.join(dl_manager.manual_dir, _OPENWEBTEXT_URLS_ZIP)
-      if not tf.io.gfile.exists(owt_path):
+      owt_path = dl_manager.manual_dir / _OPENWEBTEXT_URLS_ZIP
+      if not owt_path.exists():
         raise AssertionError(
             "For the WebText-like config, you must manually download the "
             "following file from {0} and place it in {1}: {2}".format(
@@ -606,24 +607,25 @@ def _get_pages_pcollection(self, pipeline, file_paths, dl_manager):
 
     def download_wet_file(path, dl_dir):
       url = f"{_DOWNLOAD_HOST}/{path}"
-      out_path = f"{dl_dir}/{path}"
+      out_path = epath.Path(dl_dir) / path
 
-      if tf.io.gfile.exists(out_path):
+      if out_path.exists():
         c4_utils.get_counter_inc_fn("download_wet_url")("exists")
         return out_path
 
-      tmp_dir = f"{out_path}.incomplete{uuid.uuid4().hex}"
+      tmp_dir = epath.Path(
+          f"{os.fspath(out_path)}.incomplete{uuid.uuid4().hex}"
+      )
       try:
-        tf.io.gfile.makedirs(tmp_dir)
+        tmp_dir.mkdir(parents=True, exist_ok=True)
         downloader = tfds.download.download_manager.get_downloader()
         with downloader.tqdm():
           # TODO(slebedev): Investigate why pytype infers Promise[Future[...]].
           dl_path = downloader.download(url, tmp_dir).get().path  # type: ignore
-        tf.io.gfile.rename(os.fspath(dl_path), out_path, overwrite=True)
+          dl_path = epath.Path(dl_path)
+        dl_path.rename(out_path)
       finally:
-        if tf.io.gfile.exists(tmp_dir):
-          tf.io.gfile.rmtree(tmp_dir)
-
+        tmp_dir.rmtree(missing_ok=True)
         c4_utils.get_counter_inc_fn("download_wet_url")("downloaded")
       return out_path
 
@@ -654,7 +656,7 @@ def download_wet_file(path, dl_dir):
     # Optionally filter for RealNews domains.
     # Output: [PageFeatures]
     if self.builder_config.realnewslike:
-      with tf.io.gfile.GFile(file_paths["realnews_domains"]) as f:
+      with epath.Path(file_paths["realnews_domains"]).open() as f:
         realnews_domains = json.load(f)
       pages |= beam.Filter(c4_utils.is_realnews_domain, realnews_domains)