diff --git a/simulstream/server/speech_processors/base_streamatt.py b/simulstream/server/speech_processors/base_streamatt.py
index fa7ccd4..3d31ed8 100644
--- a/simulstream/server/speech_processors/base_streamatt.py
+++ b/simulstream/server/speech_processors/base_streamatt.py
@@ -297,6 +297,27 @@ def select_text_history(self, text_history: List[str]):
         return new_history[::-1]
 
 
+class FixedCharsTextHistory:
+    """
+    Character-count-based textual history selection method that retains a pre-defined number of
+    tokens in the history (*history_chars*).
+
+    Recommended for character-level languages (e.g., Chinese, Japanese) where word-boundary
+    markers (▁) are sparse, making :class:`FixedWordsTextHistory` ineffective: when few tokens
+    carry a BOW prefix, the word counter never reaches *history_words*, so the history is never
+    trimmed and the audio history grows without bound, causing AlignAtt to cut all new tokens.
+
+    Args:
+        config (SimpleNamespace): Configuration object with an optional attribute:
+            - **history_chars (int)**: Number of tokens to retain. Defaults to 20.
+    """
+    def __init__(self, config: SimpleNamespace):
+        self.history_chars = getattr(config, "history_chars", 20)
+
+    def select_text_history(self, text_history: List[str]) -> List[str]:
+        return text_history[-self.history_chars:]
+
+
 class PunctuationTextHistory:
     """
     Punctuation textual history selection method that retains the sentence
diff --git a/uts/speech_processors/test_streamatt.py b/uts/speech_processors/test_streamatt.py
index 180c408..a20878f 100644
--- a/uts/speech_processors/test_streamatt.py
+++ b/uts/speech_processors/test_streamatt.py
@@ -15,7 +15,45 @@
 import unittest
 from types import SimpleNamespace
 
-from simulstream.server.speech_processors.base_streamatt import PunctuationTextHistory
+from simulstream.server.speech_processors.base_streamatt import (
+    FixedCharsTextHistory,
+    FixedWordsTextHistory,
+    PunctuationTextHistory,
+)
+
+
+class TestFixedWordsTextHistory(unittest.TestCase):
+    def setUp(self):
+        self.history = FixedWordsTextHistory(SimpleNamespace(history_words=3))
+
+    def test_word_level(self):
+        """ Trims history to the last 3 BOW-started words in space-separated languages. """
+        en_history = ["▁I", "▁am", "▁going", "▁to", "▁New", "▁York"]
+        self.assertEqual(
+            self.history.select_text_history(en_history),
+            ["▁to", "▁New", "▁York"])
+
+    def test_word_level_with_subwords(self):
+        """ Subword continuations (no ▁) are included in the retained word. """
+        en_history = ["▁inter", "nation", "al", "▁meet", "ing", "▁today"]
+        self.assertEqual(
+            self.history.select_text_history(en_history),
+            ["▁inter", "nation", "al", "▁meet", "ing", "▁today"])
+
+
+class TestFixedCharsTextHistory(unittest.TestCase):
+    def setUp(self):
+        self.history = FixedCharsTextHistory(SimpleNamespace(history_chars=3))
+
+    def test_char_level(self):
+        """ Keeps the last 3 tokens for character-level languages. """
+        zh_history = ['大', '家', '好', '我', '在', '谈', '论']
+        self.assertEqual(self.history.select_text_history(zh_history), ['在', '谈', '论'])
+
+    def test_shorter_than_limit(self):
+        """ Returns the full history when it is shorter than history_chars. """
+        zh_history = ['大', '家']
+        self.assertEqual(self.history.select_text_history(zh_history), ['大', '家'])
 
 
 class TestPunctuationTextHistory(unittest.TestCase):