diodiogod
e129c5314b
MAJOR: IndexTTS-2 TTS engine integration - Core IndexTTS engine with unified model interface integration - Auto-downloader for HuggingFace models + GitHub source code - Engine adapter with emotion support [Alice:angry_bob] syntax - Cache integration for IndexTTS-2 parameters - Factory registration for proper model management MAJOR: Character parser modularization (930 → 300 lines each) - Split into base_parser.py, language_resolver.py, segment_processor.py, validation.py - Added emotion support with backward compatibility - New split_by_character_with_emotions() method for advanced engines - Flexible tag parsing: [de:Alice:angry_bob] support - Language disambiguation to avoid Alice vs language code conflicts - Full backward compatibility maintained FEATURES: - IndexTTS-2 emotion disentanglement (speaker vs emotion separation) - Duration control and precise voice cloning - Multi-modal emotion control (audio refs, text, vectors) - Character switching with emotion references - All official IndexTTS-2 parameters exposed DOCS: Implementation plan and architectural TODO items added
253 lines
11 KiB
Python
253 lines
11 KiB
Python
"""
|
|
Segment Processor - Text segmentation and parsing functionality
|
|
|
|
Handles the core text processing logic for character tag parsing.
|
|
Extracted from character parser for better modularity.
|
|
"""
|
|
|
|
import re
|
|
from typing import List, Optional, Tuple, Pattern
|
|
from .base_parser import CharacterSegment
|
|
|
|
|
|
class SegmentProcessor:
|
|
"""
|
|
Handles text segmentation and character tag processing.
|
|
|
|
Responsible for:
|
|
- Line-by-line text processing
|
|
- Character tag detection and parsing
|
|
- Text segment creation with proper positioning
|
|
- Speaker format detection ("Speaker 1:", etc.)
|
|
"""
|
|
|
|
def __init__(self, character_tag_pattern: Pattern, default_character: str):
|
|
"""
|
|
Initialize segment processor.
|
|
|
|
Args:
|
|
character_tag_pattern: Regex pattern for detecting character tags
|
|
default_character: Default character for untagged text
|
|
"""
|
|
self.CHARACTER_TAG_PATTERN = character_tag_pattern
|
|
self.default_character = default_character
|
|
|
|
def parse_text_segments(self, text: str, language_resolver, character_parser) -> List[CharacterSegment]:
|
|
"""
|
|
Parse text into character-specific segments with proper line-by-line processing.
|
|
|
|
Args:
|
|
text: Input text with [Character] tags
|
|
language_resolver: LanguageResolver instance for language handling
|
|
character_parser: Main CharacterParser instance for character normalization
|
|
|
|
Returns:
|
|
List of CharacterSegment objects
|
|
"""
|
|
segments = []
|
|
|
|
# Split text by lines and process each line completely independently
|
|
lines = text.split('\n')
|
|
global_pos = 0
|
|
|
|
for line in lines:
|
|
line_start_pos = global_pos
|
|
original_line = line
|
|
line = line.strip()
|
|
|
|
if not line:
|
|
global_pos += len(original_line) + 1 # Account for empty line + newline
|
|
continue
|
|
|
|
# Each line is processed independently - no character state carries over
|
|
line_segments = self._parse_single_line(
|
|
line, line_start_pos, language_resolver, character_parser
|
|
)
|
|
segments.extend(line_segments)
|
|
|
|
global_pos += len(original_line) + 1 # +1 for newline
|
|
|
|
# If no segments were created, treat entire text as default character
|
|
if not segments and text.strip():
|
|
segments.append(CharacterSegment(
|
|
character=self.default_character,
|
|
text=text.strip(),
|
|
start_pos=0,
|
|
end_pos=len(text),
|
|
language=language_resolver.resolve_character_language(
|
|
self.default_character, None,
|
|
character_parser.character_language_defaults,
|
|
character_parser._character_language_cache,
|
|
character_parser._logged_characters
|
|
),
|
|
original_character=self.default_character,
|
|
explicit_language=False
|
|
))
|
|
|
|
return segments
|
|
|
|
def _parse_single_line(self, line: str, line_start_pos: int, language_resolver, character_parser) -> List[CharacterSegment]:
|
|
"""
|
|
Parse a single line for character tags, treating it completely independently.
|
|
|
|
Args:
|
|
line: Single line of text (no newlines)
|
|
line_start_pos: Starting position of this line in the original text
|
|
language_resolver: LanguageResolver instance
|
|
character_parser: Main CharacterParser instance
|
|
|
|
Returns:
|
|
List of CharacterSegment objects for this line only
|
|
"""
|
|
segments = []
|
|
current_pos = 0
|
|
current_character = self.default_character
|
|
current_language = language_resolver.default_language
|
|
current_emotion = None
|
|
|
|
# IMPORTANT: Each line starts fresh with narrator as default
|
|
# If the line doesn't start with a character tag, everything is narrator
|
|
|
|
# Check for manual "Speaker N:" format before processing character tags
|
|
speaker_match = self._parse_speaker_format_line(line)
|
|
if speaker_match:
|
|
speaker_name, speaker_text = speaker_match
|
|
if speaker_text.strip():
|
|
segments.append(CharacterSegment(
|
|
character=speaker_name, # Use "speaker 1", "speaker 2", etc.
|
|
text=speaker_text.strip(),
|
|
start_pos=line_start_pos,
|
|
end_pos=line_start_pos + len(line),
|
|
language=language_resolver.resolve_character_language(
|
|
speaker_name, None,
|
|
character_parser.character_language_defaults,
|
|
character_parser._character_language_cache,
|
|
character_parser._logged_characters
|
|
),
|
|
original_character=speaker_name,
|
|
explicit_language=False
|
|
))
|
|
return segments
|
|
|
|
# Quick check: if line doesn't contain any character tags, it's all narrator
|
|
if not self.CHARACTER_TAG_PATTERN.search(line):
|
|
if line.strip():
|
|
segments.append(CharacterSegment(
|
|
character=self.default_character,
|
|
text=line.strip(),
|
|
start_pos=line_start_pos,
|
|
end_pos=line_start_pos + len(line),
|
|
language=language_resolver.resolve_character_language(
|
|
self.default_character, None,
|
|
character_parser.character_language_defaults,
|
|
character_parser._character_language_cache,
|
|
character_parser._logged_characters
|
|
),
|
|
original_character=self.default_character,
|
|
explicit_language=False
|
|
))
|
|
return segments
|
|
|
|
# Find all character tags in this line
|
|
for match in self.CHARACTER_TAG_PATTERN.finditer(line):
|
|
# Add text before this tag (if any) with current character (narrator)
|
|
before_tag = line[current_pos:match.start()].strip()
|
|
if before_tag:
|
|
segments.append(CharacterSegment(
|
|
character=current_character,
|
|
text=before_tag,
|
|
start_pos=line_start_pos + current_pos,
|
|
end_pos=line_start_pos + match.start(),
|
|
language=current_language,
|
|
original_character=current_character, # Before this tag, it's already resolved
|
|
explicit_language=False, # Text before tags doesn't have explicit language
|
|
emotion=current_emotion
|
|
))
|
|
|
|
# Parse language, character, and emotion from the tag using flexible parser
|
|
raw_tag_content = match.group(1)
|
|
|
|
# Use flexible tag parser to handle emotion syntax
|
|
tag_info = language_resolver.parse_flexible_tag(raw_tag_content)
|
|
|
|
# Extract individual components
|
|
explicit_language = tag_info.get('language')
|
|
raw_character = tag_info.get('character') or self.default_character
|
|
emotion_reference = tag_info.get('emotion')
|
|
|
|
# Update current character for text after this tag
|
|
# IMPORTANT: Resolve language using original alias name before character normalization
|
|
current_language = language_resolver.resolve_character_language(
|
|
raw_character, explicit_language,
|
|
character_parser.character_language_defaults,
|
|
character_parser._character_language_cache,
|
|
character_parser._logged_characters
|
|
)
|
|
original_character = raw_character # Store original before normalization
|
|
current_explicit_language = explicit_language is not None # Track if language was explicit
|
|
|
|
# Detect language-only tags: if the original tag had empty character part and raw_character
|
|
# was defaulted to narrator, skip alias resolution to preserve narrator voice priority
|
|
is_language_only_tag = (raw_character == self.default_character and
|
|
':' in raw_tag_content and
|
|
raw_tag_content.split(':', 1)[1].strip() == '')
|
|
|
|
current_character = character_parser.normalize_character_name(
|
|
raw_character, skip_narrator_alias=is_language_only_tag
|
|
)
|
|
|
|
# Update emotion state if provided in tag
|
|
if emotion_reference is not None:
|
|
current_emotion = emotion_reference
|
|
|
|
current_pos = match.end()
|
|
|
|
# Add remaining text after last tag (or entire line if no tags)
|
|
remaining_text = line[current_pos:].strip()
|
|
if remaining_text:
|
|
segments.append(CharacterSegment(
|
|
character=current_character,
|
|
text=remaining_text,
|
|
start_pos=line_start_pos + current_pos,
|
|
end_pos=line_start_pos + len(line),
|
|
language=current_language,
|
|
original_character=original_character,
|
|
explicit_language=current_explicit_language if 'current_explicit_language' in locals() else False,
|
|
emotion=current_emotion
|
|
))
|
|
elif not segments and line.strip():
|
|
# Line with only tags and no text after - still need a segment for the line
|
|
# This handles edge cases like a line that is just "[character]"
|
|
segments.append(CharacterSegment(
|
|
character=current_character,
|
|
text="",
|
|
start_pos=line_start_pos,
|
|
end_pos=line_start_pos + len(line),
|
|
language=current_language,
|
|
original_character=original_character if 'original_character' in locals() else current_character,
|
|
explicit_language=current_explicit_language if 'current_explicit_language' in locals() else False,
|
|
emotion=current_emotion
|
|
))
|
|
|
|
return segments
|
|
|
|
def _parse_speaker_format_line(self, line: str) -> Optional[Tuple[str, str]]:
|
|
"""
|
|
Parse a line for manual "Speaker N:" format.
|
|
|
|
Args:
|
|
line: Single line to check
|
|
|
|
Returns:
|
|
Tuple of (speaker_name, text) if found, None otherwise
|
|
"""
|
|
import re
|
|
# Match "Speaker N: text" (case insensitive)
|
|
match = re.match(r'^(speaker\s*\d+)\s*:\s*(.*)$', line.strip(), re.IGNORECASE)
|
|
if match:
|
|
speaker_name = match.group(1).lower().strip() # Normalize to "speaker 1", "speaker 2", etc.
|
|
speaker_text = match.group(2)
|
|
# Normalize speaker name format
|
|
speaker_name = re.sub(r'\s+', ' ', speaker_name) # "speaker 1" -> "speaker 1"
|
|
return speaker_name, speaker_text
|
|
return None |