diff --git a/docs/changelog.md b/docs/changelog.md
index b93adb2..6959982 100644
--- a/docs/changelog.md
+++ b/docs/changelog.md
@@ -2,6 +2,40 @@
 
 ## 1.*
 
+### 1.5.0 - 24-09-02 - `str` support for HDF5
+
+Strings in hdf5 are tricky! HDF5 doesn't have native support for unicode, 
+but it can be persuaded to store data in ASCII or virtualized utf-8 under somewhat obscure conditions.
+
+This PR uses h5py's string methods to expose string datasets (compound or not) 
+via the h5proxy with the `asstr()` view method. 
+This also allows us to set strings with normal python strings,
+although hdf5 datasets can only be created with `bytes` or other non-unicode encodings.
+
+Since numpydantic isn't necessarily a tool for *creating* hdf5 files 
+(nobody should be doing that), but rather an interface to them, 
+tests are included for reading and validating (unskip the existing string tests) 
+as well as setting/getting.
+
+```python
+import h5py
+import numpy as np
+from pydantic import BaseModel
+from numpydantic import NDArray
+from typing import Any
+
+class MyModel(BaseModel):
+  array: NDArray[Any, str]
+
+h5f = h5py.File('my_data.h5', 'w')
+data = np.random.random((10,10)).astype(bytes)
+_ = h5f.create_dataset('/dataset', data=data)
+
+instance = MyModel(array=('my_data.h5', '/dataset'))
+instance[0,0] = 'hey'
+assert instance[0,0] == 'hey'
+```
+
 ### 1.4.1 - 24-09-02 - `len()` support and dunder method testing
 
 It's pretty natural to want to do `len(array)` as a shorthand for `array.shape[0]`, 
diff --git a/pyproject.toml b/pyproject.toml
index 57ae6be..59f725c 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -1,6 +1,6 @@
 [project]
 name = "numpydantic"
-version = "1.4.1"
+version = "1.5.0"
 description = "Type and shape validation and serialization for numpy arrays in pydantic models"
 authors = [
     {name = "sneakers-the-rat", email = "sneakers-the-rat@protonmail.com"},
diff --git a/src/numpydantic/interface/hdf5.py b/src/numpydantic/interface/hdf5.py
index 656273d..6bb7dda 100644
--- a/src/numpydantic/interface/hdf5.py
+++ b/src/numpydantic/interface/hdf5.py
@@ -1,5 +1,27 @@
 """
 Interfaces for HDF5 Datasets
+
+.. note::
+
+    HDF5 arrays are accessed through a proxy class :class:`.H5Proxy` .
+    Getting/setting values should work as normal, **except** that setting
+    values on nested views is impossible - 
+    
+    Specifically this doesn't work:
+    
+    .. code-block:: python
+    
+        my_model.array[0][0] = 1
+    
+    But this does work:
+    
+    .. code-block:: python
+    
+        my_model.array[0,0] = 1
+        
+    To have direct access to the hdf5 dataset, use the
+    :meth:`.H5Proxy.open` method.
+    
 """
 
 import sys
@@ -10,7 +32,7 @@ import numpy as np
 from pydantic import SerializationInfo
 
 from numpydantic.interface.interface import Interface
-from numpydantic.types import NDArrayType
+from numpydantic.types import DtypeType, NDArrayType
 
 try:
     import h5py
@@ -102,7 +124,25 @@ class H5Proxy:
         with h5py.File(self.file, "r") as h5f:
             obj = h5f.get(self.path)
             if self.field is not None:
-                obj = obj.fields(self.field)
+                if encoding := h5py.h5t.check_string_dtype(obj.dtype[self.field]):
+                    if isinstance(item, tuple):
+                        item = (*item, self.field)
+                    else:
+                        item = (item, self.field)
+
+                    try:
+                        # single string
+                        return obj[item].decode(encoding.encoding)
+                    except AttributeError:
+                        # numpy array of bytes
+                        return np.char.decode(obj[item], encoding=encoding.encoding)
+
+                else:
+                    obj = obj.fields(self.field)
+            else:
+                if h5py.h5t.check_string_dtype(obj.dtype):
+                    obj = obj.asstr()
+
             return obj[item]
 
     def __setitem__(
@@ -222,6 +262,17 @@ class H5Interface(Interface):
 
         return array
 
+    def get_dtype(self, array: NDArrayType) -> DtypeType:
+        """
+        Get the dtype from the input array
+
+        Subclasses to correctly handle
+        """
+        if h5py.h5t.check_string_dtype(array.dtype):
+            return str
+        else:
+            return array.dtype
+
     @classmethod
     def to_json(cls, array: H5Proxy, info: Optional[SerializationInfo] = None) -> dict:
         """
diff --git a/src/numpydantic/interface/interface.py b/src/numpydantic/interface/interface.py
index 3dc3fdc..1ef307f 100644
--- a/src/numpydantic/interface/interface.py
+++ b/src/numpydantic/interface/interface.py
@@ -126,7 +126,10 @@ class Interface(ABC, Generic[T]):
         if isinstance(self.dtype, tuple):
             valid = dtype in self.dtype
         elif self.dtype is np.str_:
-            valid = getattr(dtype, "type", None) is np.str_ or dtype is np.str_
+            valid = getattr(dtype, "type", None) in (np.str_, str) or dtype in (
+                np.str_,
+                str,
+            )
         else:
             # try to match as any subclass, if self.dtype is a class
             try:
diff --git a/tests/fixtures.py b/tests/fixtures.py
index 6fbc5ad..9d5bba6 100644
--- a/tests/fixtures.py
+++ b/tests/fixtures.py
@@ -122,13 +122,22 @@ def hdf5_array(
         compound: bool = False,
     ) -> H5ArrayPath:
         array_path = "/" + "_".join([str(s) for s in shape]) + "__" + dtype.__name__
+
         if not compound:
-            data = np.random.random(shape).astype(dtype)
+            if dtype is str:
+                data = np.random.random(shape).astype(bytes)
+            else:
+                data = np.random.random(shape).astype(dtype)
             _ = hdf5_file.create_dataset(array_path, data=data)
             return H5ArrayPath(Path(hdf5_file.filename), array_path)
         else:
-            dt = np.dtype([("data", dtype), ("extra", "i8")])
-            data = np.zeros(shape, dtype=dt)
+
+            if dtype is str:
+                dt = np.dtype([("data", np.dtype("S10")), ("extra", "i8")])
+                data = np.array([("hey", 0)] * np.prod(shape), dtype=dt).reshape(shape)
+            else:
+                dt = np.dtype([("data", dtype), ("extra", "i8")])
+                data = np.zeros(shape, dtype=dt)
             _ = hdf5_file.create_dataset(array_path, data=data)
             return H5ArrayPath(Path(hdf5_file.filename), array_path, "data")
 
diff --git a/tests/test_interface/test_hdf5.py b/tests/test_interface/test_hdf5.py
index 23d26a3..891dd9f 100644
--- a/tests/test_interface/test_hdf5.py
+++ b/tests/test_interface/test_hdf5.py
@@ -78,8 +78,6 @@ def test_hdf5_shape(shape_cases, hdf5_array, compound):
 
 @pytest.mark.parametrize("compound", [True, False])
 def test_hdf5_dtype(dtype_cases, hdf5_array, compound):
-    if dtype_cases.dtype is str:
-        pytest.skip("hdf5 cant do string arrays")
     _test_hdf5_case(dtype_cases, hdf5_array, compound)
 
 
@@ -157,3 +155,22 @@ def test_compound_dtype(tmp_path):
     assert all(instance.array[1, :] == 0)
     instance.array[1] = 2
     assert all(instance.array[1] == 2)
+
+
+@pytest.mark.parametrize("compound", [True, False])
+def test_strings(hdf5_array, compound):
+    """
+    HDF5 proxy can get and set strings just like any other dtype
+    """
+    array = hdf5_array((10, 10), str, compound=compound)
+
+    class MyModel(BaseModel):
+        array: NDArray[Shape["10, 10"], str]
+
+    instance = MyModel(array=array)
+    instance.array[0, 0] = "hey"
+    assert instance.array[0, 0] == "hey"
+    assert isinstance(instance.array[0, 1], str)
+
+    instance.array[1] = "sup"
+    assert all(instance.array[1] == "sup")