Add custom dict

README.md

@@ -142,6 +142,49 @@ print(eng("project"))                           # Output: /proː˧.t͡ɕek̚˨˩
 ```
 
 
+### Custom Word Pronunciation Dictionary
+
+PyThaiTTS supports custom word dictionaries to define custom Thai pronunciation (as Thai text reading or IPA phonemes) for specific words, acronyms, or names.
+
+Dictionaries can be passed as Python dicts, JSON files, or delimited text files (TSV, CSV, or colon-separated):
+
+```python
+from pythaitts import TTS
+
+# 1. Provide custom Thai pronunciation (hyphenated or plain text)
+custom_dict = {
+    "เพชรบุรี": "เพ็ด-ชะ-บุ-รี",
+    "สถิติ": "สะ-ถิ-ติ",
+    "ChatGPT": "แชตจีพีที",
+}
+
+tts = TTS(custom_dict=custom_dict)
+tts.tts("ยินดีต้อนรับสู่เพชรบุรี", filename="custom.wav")
+
+# 2. Or pass per-call or with IPA phonemes directly
+tts.tts(
+    "ผมไปเพชรบุรี",
+    custom_dict={"เพชรบุรี": "/pʰet̚˨˩.t͡ɕʰa˦˥.bu˨˩.riː˧/"},
+    filename="ipa.wav",
+)
+
+# 3. Add or update custom words dynamically
+tts.add_custom_dict({"ก.พ.": "กอ-พอ"})
+
+# 4. Load from a JSON or TSV file
+tts.set_custom_dict("my_words.json")
+```
+
+You can also use custom dictionaries with G2P directly:
+
+```python
+from pythaitts.pretrained.fastthaig2p import G2P
+
+g2p = G2P(custom_dict={"เพชรบุรี": "เพ็ด-ชะ-บุ-รี"})
+print(g2p.convert("ผมไปเพชรบุรี"))
+```
+
+
 ### Text Preprocessing
 
 PyThaiTTS includes automatic text preprocessing to improve TTS quality:

demo.py

@@ -41,6 +41,13 @@ def main():
         print(f"✓ Waveform generated successfully!")
         print(f"Waveform shape: {waveform.shape}")
         print()
+
+        # Demonstrate custom word pronunciation dictionary
+        print("Demonstrating custom word pronunciation dictionary...")
+        custom_dict = {"พายไทยทีทีเอส": "พาย-ไทย-ที-ที-เอส"}
+        waveform_custom = tts.tts(text, custom_dict=custom_dict, return_type="waveform")
+        print(f"✓ Custom pronunciation waveform generated! Shape: {waveform_custom.shape}")
+        print()
 
         print("=" * 60)
         print("Demo completed successfully!")

pythaitts/__init__.py

@@ -4,9 +4,15 @@
 """
 __version__ = "0.8.0"
 
-from typing import Union, Iterable, Iterator
+from typing import Union, Iterable, Iterator, Optional, Dict, Any
+from pathlib import Path
 
-from pythaitts.preprocess import preprocess_text, num_to_thai, expand_maiyamok
+from pythaitts.preprocess import (
+    preprocess_text,
+    num_to_thai,
+    expand_maiyamok,
+    load_custom_dict,
+)
 from pythaitts.pretrained.fastthaig2p.en_g2p import (
     EnglishG2P,
     en_g2p,
@@ -24,12 +30,21 @@
 
 
 class TTS:
-    def __init__(self, pretrained="fastthaig2p", mode="last_checkpoint", version="1.0", device:str="cpu", **kwargs) -> None:
+    def __init__(
+        self,
+        pretrained="fastthaig2p",
+        mode="last_checkpoint",
+        version="1.0",
+        device: str = "cpu",
+        custom_dict: Optional[Union[Dict[str, str], str, Path, Iterable[Any]]] = None,
+        **kwargs,
+    ) -> None:
         """
         :param str pretrained: TTS pretrained (fastthaig2p, lunarlist_onnx, khanomtan, lunarlist, vachana)
         :param str mode: pretrained mode (fastthaig2p, lunarlist_onnx, and vachana don't support)
         :param str version: model version (default is 1.0 or 1.1)
         :param str device: device for running model. (fastthaig2p, lunarlist_onnx, and vachana support CPU only.)
+        :param custom_dict: Optional custom word dictionary for custom Thai pronunciation in words
 
         **Options for mode**
             * *last_checkpoint* (default) - last checkpoint of model
@@ -54,6 +69,9 @@ def __init__(self, pretrained="fastthaig2p", mode="last_checkpoint", version="1.
         self.pretrained = pretrained
         self.mode = mode
         self.device = device
+        self.custom_dict = custom_dict
+        if custom_dict is not None:
+            kwargs["custom_dict"] = custom_dict
         self.load_pretrained(version=version, **kwargs)
 
     def load_pretrained(self, version="1.0", **kwargs):
@@ -80,7 +98,40 @@ def load_pretrained(self, version="1.0", **kwargs):
                 "PyThaiTTS doesn't support %s pretrained." % self.pretrained
             )
 
-    def tts(self, text: str, speaker_idx: str = "thai_som", language_idx: str = "th-th", return_type: str = "file", filename: str = None, preprocess: bool = True, **kwargs):
+    def add_custom_dict(
+        self, custom_dict: Union[Dict[str, str], str, Path, Iterable[Any]]
+    ) -> None:
+        """Add custom word pronunciations to the TTS model."""
+        loaded = load_custom_dict(custom_dict)
+        if self.custom_dict is None:
+            self.custom_dict = {}
+        elif isinstance(self.custom_dict, (str, Path)):
+            self.custom_dict = load_custom_dict(self.custom_dict)
+        elif not isinstance(self.custom_dict, dict):
+            self.custom_dict = load_custom_dict(self.custom_dict)
+        self.custom_dict.update(loaded)
+        if hasattr(self.model, "add_custom_dict"):
+            self.model.add_custom_dict(loaded)
+
+    def set_custom_dict(
+        self, custom_dict: Union[Dict[str, str], str, Path, Iterable[Any]]
+    ) -> None:
+        """Replace custom word pronunciations with new dictionary."""
+        self.custom_dict = custom_dict
+        if hasattr(self.model, "set_custom_dict"):
+            self.model.set_custom_dict(custom_dict)
+
+    def tts(
+        self,
+        text: str,
+        speaker_idx: str = "thai_som",
+        language_idx: str = "th-th",
+        return_type: str = "file",
+        filename: str = None,
+        preprocess: bool = True,
+        custom_dict: Optional[Union[Dict[str, str], str, Path, Iterable[Any]]] = None,
+        **kwargs,
+    ):
         """
         speech synthesis
 
@@ -90,31 +141,64 @@ def tts(self, text: str, speaker_idx: str = "thai_som", language_idx: str = "th-
         :param str return_type: return type (default is file)
         :param str filename: path filename for save wav file if return_type is file.
         :param bool preprocess: whether to preprocess text (convert numbers to Thai text and expand ๆ). Default is True.
+        :param custom_dict: Optional custom word dictionary for custom Thai pronunciation in words
         :param kwargs: Additional parameters passed to the underlying model.
         """
+        effective_custom_dict = (
+            custom_dict if custom_dict is not None else getattr(self, "custom_dict", None)
+        )
+
         # Preprocess text if requested
         if preprocess:
             from pythaitts.preprocess import preprocess_text
-            text = preprocess_text(text)
+
+            if self.pretrained not in ("fastthaig2p", "FastThaiG2P") and effective_custom_dict:
+                text = preprocess_text(text, custom_dict=effective_custom_dict)
+            else:
+                text = preprocess_text(text)
+        elif self.pretrained not in ("fastthaig2p", "FastThaiG2P") and effective_custom_dict:
+            from pythaitts.preprocess import preprocess_text
+
+            text = preprocess_text(
+                text,
+                expand_numbers=False,
+                expand_maiyamok_char=False,
+                custom_dict=effective_custom_dict,
+            )
 
         if self.pretrained == "lunarlist" or self.pretrained == "lunarlist_onnx":
-            return self.model(text=text,return_type=return_type,filename=filename)
+            return self.model(text=text, return_type=return_type, filename=filename)
         elif self.pretrained == "vachana":
             if speaker_idx in ("thai_som", "Linda", None):
                 speaker_idx = "th_f_1"
-            return self.model(text=text,speaker_idx=speaker_idx,return_type=return_type,filename=filename, **kwargs)
+            return self.model(
+                text=text,
+                speaker_idx=speaker_idx,
+                return_type=return_type,
+                filename=filename,
+                **kwargs,
+            )
         elif self.pretrained in ("fastthaig2p", "FastThaiG2P"):
             if speaker_idx in ("Linda", None):
                 speaker_idx = "thai_som"
-            return self.model(text=text,speaker_idx=speaker_idx,return_type=return_type,filename=filename, **kwargs)
+            call_kwargs = dict(kwargs)
+            if effective_custom_dict is not None:
+                call_kwargs["custom_dict"] = effective_custom_dict
+            return self.model(
+                text=text,
+                speaker_idx=speaker_idx,
+                return_type=return_type,
+                filename=filename,
+                **call_kwargs,
+            )
         if speaker_idx in ("thai_som", None):
             speaker_idx = "Linda"
         return self.model(
             text=text,
             speaker_idx=speaker_idx,
             language_idx=language_idx,
             return_type=return_type,
-            filename=filename
+            filename=filename,
         )
 
     def stream(
@@ -125,6 +209,7 @@ def stream(
         play: bool = False,
         preprocess: bool = True,
         max_phonemes: int = 400,
+        custom_dict: Optional[Union[Dict[str, str], str, Path, Iterable[Any]]] = None,
         **kwargs,
     ):
         """
@@ -136,9 +221,14 @@ def stream(
         :param bool play: Whether to play audio chunks in real-time to speakers
         :param bool preprocess: Whether to preprocess text (numbers to words, ๆ)
         :param int max_phonemes: Maximum phonemes per synthesized chunk
+        :param custom_dict: Optional custom word dictionary for custom Thai pronunciation in words
         :param kwargs: Additional parameters passed to the model
         :return: Generator yielding audio chunks
         """
+        effective_custom_dict = (
+            custom_dict if custom_dict is not None else getattr(self, "custom_dict", None)
+        )
+
         if self.pretrained in ("fastthaig2p", "FastThaiG2P"):
             if speaker_idx in ("Linda", None):
                 speaker_idx = "thai_som"
@@ -149,20 +239,25 @@ def stream(
                 play=play,
                 preprocess=preprocess,
                 max_phonemes=max_phonemes,
+                custom_dict=effective_custom_dict,
                 **kwargs,
             )
         else:
             from pythaitts.realtime import stream_text_to_chunks
 
             def _gen():
                 for chunk in stream_text_to_chunks(
-                    text, max_phonemes=max_phonemes, preprocess=preprocess
+                    text,
+                    max_phonemes=max_phonemes,
+                    preprocess=preprocess,
+                    custom_dict=effective_custom_dict,
                 ):
                     yield self.tts(
                         text=chunk,
                         speaker_idx=speaker_idx,
                         return_type=return_type,
                         preprocess=False,
+                        custom_dict=effective_custom_dict,
                         **kwargs,
                     )
 
@@ -182,6 +277,7 @@ def _gen():
     "en_to_ipa",
     "en_ipa_to_thai_ipa",
     "preprocess_text",
+    "load_custom_dict",
     "num_to_thai",
     "expand_maiyamok",
     "chunk_text",