diff --git a/src/parallel_corpus/graph.py b/src/parallel_corpus/graph.py
index fdbeb1e..b29ebeb 100644
--- a/src/parallel_corpus/graph.py
+++ b/src/parallel_corpus/graph.py
@@ -4,7 +4,7 @@
 import logging
 import re
 from dataclasses import dataclass
-from typing import Dict, Iterable, List, Optional, TypeVar
+from typing import Dict, Iterable, List, Optional, TypedDict, TypeVar
 
 import parallel_corpus.shared.ranges
 import parallel_corpus.shared.str_map
@@ -87,6 +87,12 @@ def init(s: str, *, manual: bool = False) -> Graph:
     return init_from(token.tokenize(s), manual=manual)
 
 
+def init_with_source_and_target(source: str, target: str, *, manual: bool = False) -> Graph:
+    return init_from_source_and_target(
+        source=token.tokenize(source), target=token.tokenize(target), manual=manual
+    )
+
+
 def init_from(tokens: List[str], *, manual: bool = False) -> Graph:
     return align(
         Graph(
@@ -99,10 +105,59 @@ def init_from(tokens: List[str], *, manual: bool = False) -> Graph:
     )
 
 
+def init_from_source_and_target(
+    source: List[str], target: List[str], *, manual: bool = False
+) -> Graph:
+    source_tokens = token.identify(source, "s")
+    target_tokens = token.identify(target, "t")
+    return align(
+        Graph(
+            source=source_tokens,
+            target=target_tokens,
+            edges=edge_record(
+                itertools.chain(
+                    (edge([s.id], [], manual=manual) for s in source_tokens),
+                    (edge([t.id], [], manual=manual) for t in target_tokens),
+                )
+            ),
+        )
+    )
+
+
+class TextLabels(TypedDict):
+    text: str
+    labels: List[str]
+
+
+def from_unaligned(st: SourceTarget[List[TextLabels]]) -> Graph:
+    """Initialize a graph from unaligned tokens"""
+    edges: Dict[str, Edge] = {}
+
+    def proto_token_to_token(tok: TextLabels, i: int, prefix: str) -> Token:
+        id_ = f"{prefix}{i}"
+        e = edge([id_], tok["labels"], manual=False)
+        edges[id_] = e
+        return Token(tok["text"], id_)
+
+    def proto_tokens_to_tokens(toks: List[TextLabels], side: Side) -> List[Token]:
+        return [
+            proto_token_to_token(tok, i, "s" if side == Side.source else "t")
+            for i, tok in enumerate(toks)
+        ]
+
+    g = map_sides(st, proto_tokens_to_tokens)
+
+    return align(Graph(source=g.source, target=g.target, edges=edges))
+
+
 def modify(g: Graph, from_: int, to: int, text: str, side: Side = Side.target) -> Graph:
     return align(unaligned_modify(g, from_, to, text, side))
 
 
+def set_source(g: Graph, text: str) -> Graph:
+    return align(unaligned_set_side(g, Side.source, text))
+
+
 def set_target(g: Graph, text: str) -> Graph:
     return align(unaligned_set_side(g, Side.target, text))
 
diff --git a/src/parallel_corpus/py.typed b/src/parallel_corpus/py.typed
new file mode 100644
index 0000000..e69de29
diff --git a/tests/test_graph.py b/tests/test_graph.py
index 8726e4b..5103be0 100644
--- a/tests/test_graph.py
+++ b/tests/test_graph.py
@@ -2,7 +2,7 @@
 
 import pytest
 from parallel_corpus import graph, token
-from parallel_corpus.source_target import Side
+from parallel_corpus.source_target import Side, SourceTarget
 
 
 def test_graph_init() -> None:
@@ -16,6 +16,44 @@ def test_graph_init() -> None:
     assert g.edges == edges
 
 
+def test_init_from_source_and_target_1() -> None:
+    g = graph.init_with_source_and_target(source="apa", target="apa")
+    assert g == graph.init("apa")
+
+
+def test_init_from_source_and_target_2() -> None:
+    g = graph.init_with_source_and_target(source="apa bepa", target="apa")
+    expected_source = token.identify(token.tokenize("apa bepa"), "s")
+    expected_target = token.identify(token.tokenize("apa"), "t")
+    g_expected = graph.Graph(
+        source=expected_source,
+        target=expected_target,
+        edges=graph.edge_record([graph.edge(["s0", "t0"], []), graph.edge(["s1"], [])]),
+    )
+    assert g == g_expected
+
+
+def test_init_from_source_and_target_3() -> None:
+    g = graph.init_with_source_and_target(source="apa", target="bepa apa")
+    expected_source = token.identify(token.tokenize("apa"), "s")
+    expected_target = token.identify(token.tokenize("bepa apa"), "t")
+    g_expected = graph.Graph(
+        source=expected_source,
+        target=expected_target,
+        edges=graph.edge_record([graph.edge(["s0", "t1"], []), graph.edge(["t0"], [])]),
+    )
+    assert g == g_expected
+
+
+def test_from_unaligned() -> None:
+    g = graph.from_unaligned(
+        SourceTarget(
+            source=[{"text": "apa ", "labels": []}], target=[{"text": "apa ", "labels": []}]
+        )
+    )
+    assert g == graph.init("apa")
+
+
 def test_graph_case1() -> None:
     first = "Jonathan saknades , emedan han , med sin vapendragare , redan på annat håll sökt och anträffat fienden ."  # noqa: E501
     second = "Jonat han saknades , emedan han , med sin vapendragare , redan på annat håll sökt och anträffat fienden ."  # noqa: E501
@@ -38,6 +76,17 @@ def test_graph_case2() -> None:
     assert "e-s0-s1-t20" in gm.edges
 
 
+def test_set_source() -> None:
+    source = "Jonat han saknades"
+    target = "Jonathan saknaes"
+
+    g = graph.init(target)
+
+    gm = graph.set_source(g, source)
+    print(f"{gm=}")
+    assert "e-s2-s3-t0" in gm.edges
+
+
 def test_unaligned_set_side() -> None:
     g0 = graph.init("a bc d")
     print(">>> test_unaligned_set_side")