AI Text-to-Speech Engine with Content-Based Style Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional text-to-speech services cannot reflect various speech styles and lack the ability to generate texts with styles based on content, resulting in uniform speech styles for texts from the same user.
Innovation Solution
An artificial intelligence apparatus that extracts content keywords from input texts, determines appropriate speech or text styles, and uses TTS or text generation engines to generate output accordingly, allowing for varied styles at the unit level such as words or clauses.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a predetermined Text-To-Speech engine is used, then the conversion process is simple, but the speech style cannot reflect various content styles
Solution Approach 1:
The patent segments the text processing into multiple stages: keyword extraction, style determination, and TTS generation. By dividing the complex task of style-aware speech generation into separate functional modules, the system achieves versatility in speech styles while maintaining manageable complexity through modular architecture.
Solution Approach 2:
The patent introduces an intermediary style determination module that acts as a mediator between the input text and the TTS engine. This intermediary component analyzes text content, determines appropriate styles, and selects corresponding TTS engines, enabling flexible speech style adaptation without directly complicating the core TTS conversion process.
2Adaptability or versatility
If uniform TTS engine is used for all texts, then the processing is consistent and simple, but it cannot generate different styles for different content
Solution Approach 1:
The patent applies preliminary action by extracting keywords and determining styles before the actual TTS generation. This pre-processing step analyzes the text content in advance, identifies appropriate speech styles, and prepares the TTS engine selection, thereby enabling content-style matching while maintaining efficient processing through structured preparation.
Solution Approach 2:
The patent implements dynamics by making the TTS engine selection adaptive rather than fixed. The system dynamically chooses different TTS engines based on the determined speech style, allowing the processing pipeline to adjust its behavior according to content requirements while maintaining consistent overall processing through automated decision-making.
3Adaptability or versatility
If the same speech style is applied to all texts from a user, then the processing is uniform and efficient, but it cannot reflect emotional nuances of different content
Solution Approach 1:
The patent applies local quality by determining speech styles at the sentence or clause level rather than applying a uniform style to all user texts. This allows different emotional nuances to be captured in different parts of the text, with each local unit receiving appropriate style treatment while the overall system remains manageable through localized analysis.
Solution Approach 2:
The patent utilizes parameter changes by varying speech style parameters (such as tone, pitch, speed) based on the determined style for each text unit. This enables emotional expression through parameter variation while keeping the determination process efficient by using keyword-based style classification rather than complex real-time analysis.
4Adaptability or versatility
If style is determined at the sentence level, then the processing is simple, but it cannot emphasize specific units within sentences
Solution Approach 1:
The patent segments the text into smaller units (sentences, clauses, or words) and applies style determination to each unit independently. This segmentation enables unit-level style variation and emphasis while maintaining processing simplicity through standardized treatment of each segment, avoiding the need for complex holistic analysis.
Data Source
AI summary
Provided is an artificial intelligence (AI) apparatus for generating a speech having a content-based style, including: a memory configured to store a plurality of TTS (Text-To-Speech) engines; and a processor configured to: obtain image data or text data containing a text, extract at least one content keyword corresponding to the text, determine a speech style based on the extracted content keyword, generate a speech corresponding to the text by using a TTS engine corresponding to the determined speech style among the plurality of TTS engines, and output the generated speech.