From 82c2a6eaa19fa2ccad24399bf14b8d42b6d83e18 Mon Sep 17 00:00:00 2001
From: david-bowiegxw <834563048@qq.com>
Date: Tue, 18 Aug 2026 15:43:57 +0900
Subject: [PATCH] fix(webui): preview IndexTTS 2.5 text segments (#777)

---
 webui.py | 29 +++++++++++++++++++++++------
 1 file changed, 23 insertions(+), 6 deletions(-)

diff --git a/webui.py b/webui.py
index f595f8e..bfc7ab2 100644
--- a/webui.py
+++ b/webui.py
@@ -1047,12 +1047,22 @@ with gr.Blocks(
                         outputs=example_outputs
     )
 
-    def on_input_text_change(text, max_text_tokens_per_segment):
+    def on_input_text_change(text, max_text_tokens_per_segment, lang_choice):
         if text and len(text) > 0:
             if IS_V25:
-                # v2.5 uses tiktoken encoder, no split_segments; show token count only
-                tokens = tts.tokenizer.encode(text, allowed_special='all')
-                data = [[0, text, len(tokens)]]
+                lang_prefix = f'<|{(lang_choice or "ZH").lower()}|> '
+                segments = tts.split_text_by_tokens(
+                    text,
+                    int(max_text_tokens_per_segment),
+                    lang_prefix,
+                )
+                data = []
+                for i, segment_str in enumerate(segments):
+                    tokens_count = len(tts.tokenizer.encode(
+                        lang_prefix + segment_str,
+                        allowed_special='all',
+                    ))
+                    data.append([i, segment_str, tokens_count])
             else:
                 text_tokens_list = tts.tokenizer.tokenize(text)
                 segments = tts.tokenizer.split_segments(text_tokens_list, max_text_tokens_per_segment=int(max_text_tokens_per_segment))
@@ -1186,16 +1196,23 @@ with gr.Blocks(
 
     input_text_single.change(
         on_input_text_change,
-        inputs=[input_text_single, max_text_tokens_per_segment],
+        inputs=[input_text_single, max_text_tokens_per_segment, lang_dropdown],
         outputs=[segments_preview]
     )
 
     max_text_tokens_per_segment.change(
         on_input_text_change,
-        inputs=[input_text_single, max_text_tokens_per_segment],
+        inputs=[input_text_single, max_text_tokens_per_segment, lang_dropdown],
         outputs=[segments_preview]
     )
 
+    if IS_V25:
+        lang_dropdown.change(
+            on_input_text_change,
+            inputs=[input_text_single, max_text_tokens_per_segment, lang_dropdown],
+            outputs=[segments_preview]
+        )
+
     prompt_audio.upload(update_prompt_audio,
                          inputs=[],
                          outputs=[gen_button])