Add Hungarian number to words converter

This commit introduces a new module `num2words_hu.py` that implements a number to words converter for the Hungarian language. The module includes functionality for converting cardinal numbers, ordinal numbers, years, and currency values into their Hungarian verbal representations. It builds upon a base class for European number conversion and includes specific adaptations for Hungarian linguistic rules and currency formats.

Author: sarpba

README.md

@@ -91,7 +91,7 @@ pip install num2words
 > 
 > ```bash
 > git clone https://github.com/SWivid/F5-TTS.git
-> cd F5-TTS
+> cd F5-TTS_hun
 > # git submodule update --init --recursive  # (optional, if need > bigvgan)
 > pip install -e .
 > ```

src/f5_tts/infer/infer_gradio.py

@@ -1,9 +1,6 @@
-# ruff: noqa: E402
-# Above allows ruff to ignore E402: module level import not at top of file
-
 import gc
 import json
-import os  # needed for normaliser folder reading
+import os
 import re
 import tempfile
 from collections import OrderedDict
@@ -46,6 +43,12 @@ def gpu_decorator(func):
 from f5_tts.model import DiT, UNetT
 
 
+# Define the absolute path to the normalisers directory relative to this script.
+# This makes the script independent of the current working directory.
+SCRIPT_DIR = os.path.dirname(os.path.abspath(__file__))
+NORMALISERS_DIR = os.path.join(SCRIPT_DIR, "normalisers")
+
+
 DEFAULT_TTS_MODEL = "F5-TTS_v1"
 tts_model_choice = DEFAULT_TTS_MODEL
 
@@ -58,14 +61,13 @@ def gpu_decorator(func):
 
 def get_normaliser_choices():
     """
-    Scan the ./normalisers folder for subfolders that contain a normaliser.py file.
-    Returns a list of possible normaliser choices.
+    Scan the normalisers folder (relative to this script) for subfolders
+    that contain a normaliser.py file. Returns a list of possible normaliser choices.
     """
-    normaliser_path = "./normalisers"
     choices = []
-    if os.path.exists(normaliser_path) and os.path.isdir(normaliser_path):
-        for item in os.listdir(normaliser_path):
-            subdir = os.path.join(normaliser_path, item)
+    if os.path.exists(NORMALISERS_DIR) and os.path.isdir(NORMALISERS_DIR):
+        for item in os.listdir(NORMALISERS_DIR):
+            subdir = os.path.join(NORMALISERS_DIR, item)
             if os.path.isdir(subdir) and os.path.exists(os.path.join(subdir, "normaliser.py")):
                 choices.append(item)
     return choices
@@ -315,7 +317,8 @@ def basic_tts(
         # -- Normaliser step --
         processed_text = gen_text_input
         if normaliser_choice_input != "None":
-            normaliser_file = os.path.join("normalisers", normaliser_choice_input, "normaliser.py")
+            # Use the absolute path defined at the start of the script
+            normaliser_file = os.path.join(NORMALISERS_DIR, normaliser_choice_input, "normaliser.py")
             if os.path.exists(normaliser_file):
                 try:
                     import importlib.util
@@ -673,7 +676,8 @@ def generate_multistyle_speech(
         # -- Load normaliser module if chosen --
         normaliser_module = None
         if normaliser_choice != "None":
-            normaliser_file = os.path.join("normalisers", normaliser_choice, "normaliser.py")
+            # Use the absolute path defined at the start of the script
+            normaliser_file = os.path.join(NORMALISERS_DIR, normaliser_choice, "normaliser.py")
             if os.path.exists(normaliser_file):
                 try:
                     import importlib.util
@@ -1072,7 +1076,7 @@ def chat_pipeline(conv_state, audio_path, text_msg, system_prompt, ref_audio, re
 with gr.Blocks() as app:
     gr.Markdown(
         f"""
-# E2/F5 TTS
+# E2/F5 TTS_hun normalizálóval ellátott változata
 
 This is {"a local web UI for [F5 TTS](https://github.com/SWivid/F5-TTS)" if not USING_SPACES else "an online demo for [F5-TTS](https://github.com/SWivid/F5-TTS)"} with advanced batch processing support. This app supports the following TTS models:
 
@@ -1187,8 +1191,8 @@ def set_custom_model(custom_ckpt_path, custom_vocab_path, custom_model_cfg_str):
     global_choose_normaliser = gr.Radio(
         choices=["None"] + normaliser_choices_list,
         label="Choose Text Normaliser (Global for Basic/Multi-Speech)",
-        value="None",
-        info="Applies to text in 'Basic-TTS' and 'Multi-Speech' tabs. Requires normaliser.py in ./normalisers/<name>/ folder."
+        value="hun" if "hun" in normaliser_choices_list else "None",
+        info="Applies to text in 'Basic-TTS' and 'Multi-Speech' tabs. Looks for normaliser.py in <script_dir>/normalisers/<name>/ folder."
     )
 
 
@@ -1312,4 +1316,4 @@ def main(port, host, share, api, root_path, inbrowser):
     if not USING_SPACES:
         main()
     else:
-        app.queue().launch()
+        app.queue().launch()

src/f5_tts/infer/normalisers/hun/changes.csv

@@ -29,6 +29,8 @@ geneieve, dzsenív
 boimler, bojler
 one, egy
 two, kettő
+microsoft, mikroszoft
+malware, malvér
 gpu, gépéú
 openai, openéáj
 deepseek, dípszík
@@ -753,3 +755,120 @@ Fidelis,Fidélisz
 Gót,Gót
 Kobra,Kobra
 Kurtz,Kurc
+FBI,efbíáj
+CIA,sziáéj
+NASA,náza
+UN,uenn
+EU,eu
+NATO,nátó
+USA,júeszéj
+UK,júkéj
+UFO,ufó
+NBA,enbíéj
+NFL,enefel
+BBC,bíbiszí
+CNN,szíenen
+IT,ájti
+AI,éjáj
+CEO,szííó
+GPS,dzsípíesz
+LED,led
+RAM,ram
+ROM,rom
+WWW, vévévé
+ZIP,zip
+PIN,pin
+ID,ájdí
+LAN,lan
+WiFi,vifi
+ASAP,éjszep
+DIY,doitjuorszelf
+FAQ,fakú
+BRB,bíárbí
+BTW,bájthevéj
+LOL,lol
+OMG,óemdzsi
+IDK,ájdíkéj
+IQ,ikú
+EQ,ekú
+PR,píár
+GIF,gif
+JPEG,jpeg
+CSGO,cségó
+MAC,mek
+PC,pécé
+BIOS,biosz
+DOS,dosz
+GUI,gui
+SATA,sata
+RAID,réd
+JSON,dzséjszon
+XML,ikszemel
+API,api
+PhD,pí-ejcs-dí
+CV,szíví
+CFO,szíefó
+COO,szíóó
+CTO,szítíó
+R&D,áren-dí
+B2B,bí-tú-bí
+B2C,bí-tú-szí
+CRM,szíárem
+ERP,íárpí
+KPI,kéjpíáj
+OKR,óukéjár
+ROI,ároj
+VoIP,vóip
+CAD,ked
+CAM,kem
+CAE,szíéjí
+SCM,esz-szí-em
+BPM,bí-pí-em
+IPO,ájpíó
+ETF,í-tí-ef
+GDP,dzsí-dí-pí
+VAT,víéjtí
+IRS,ájáresz
+DEA,dííéj
+NSA,eneszéj
+DHS,díéjcsesz
+DoD,díoudí
+AFK,éjefkéj
+NPC,enpíszí
+RPG,árpídzsí
+HUD,hád
+DLC,díelszí
+SDK,eszdíkéj
+DBMS,díbíemesz
+IoT,ájótí
+ML,emel
+DL,díel
+NLP,enelpí
+DDR,dídídár
+PSU,píeszjú
+CDN,szídíen
+TOS,tíóesz
+EULA,júlá
+GDPR,dzsídípíár
+HIPAA,hipá
+DMCA,díemszéj
+ACL,éjszíel
+IPSec,ájpíszek
+SQLi,esz-kjú-el-áj
+XSS,ekszesz
+CSRF,szíeszáref
+MITM,emájtíem
+ICU,ájszjú
+CPR,szípíár
+HIV,hív
+AIDS,éjdz
+DNA,díéjenéj
+RNA,árenéj
+BMI,bíemáj
+CDC,szídíszí
+EMT,íemtí
+MRSA,emerszá
+MRI,emáráj
+CT,szítí
+X-ray,ekszréj
+ADHD,éjdíéjcsdí