Text-to-Speech Terminal Switching for Reduced Playback Delay

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

The existing text-to-speech (TTS) technology experiences delays in audio stream delivery from servers, leading to prolonged playback times and negatively impacting user experience in online synthesis processes.

Innovation Solution

Implementing a hybrid synthesis approach where the terminal switches between online and offline methods based on connection success and text content, using offline synthesis if the first frame of playable audio data is not received within a preset duration.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If online synthesis is used, then speech can be generated remotely with potentially better quality, but playback delay increases due to server processing time

Engineering Contradiction:
Improvespeech generation qualityVSAvoidplayback delay
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The system dynamically switches between online and offline synthesis methods based on real-time conditions. The terminal determines whether to use online synthesis by checking if audio data can be received within a preset duration, and switches to offline synthesis when online method fails to meet timing requirements, creating a dynamic adaptive system

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The system changes the synthesis method parameter based on timing constraints. When the server cannot return audio data within the preset duration, the system changes from online synthesis parameter to offline synthesis parameter, allowing immediate local playback without further server dependency

Inventive Principle:
Principle #35Parameter changes

2Loss of time

If offline synthesis is used, then playback delay is reduced through local processing, but storage requirements increase due to need to store speech for massive words

Engineering Contradiction:
Improveplayback delayVSAvoidstorage capacity
Core Design Contradiction:
Loss of timeVSQuantity of substance

Solution Approach 1:

Instead of storing complete speech files for all possible words (excessive action), the system stores only necessary audio data segments locally for offline synthesis when needed (partial action). This selective approach reduces storage requirements compared to storing all possible speech content

Inventive Principle:
Principle #16Partial or excessive action

3Device complexity

If server synthesis is used, then processing power requirements are reduced at terminal, but connection dependency increases and may fail under network constraints

Engineering Contradiction:
Improveterminal processing requirementVSAvoidconnection dependency
Core Design Contradiction:
Device complexityVSReliability

Solution Approach 1:

The terminal prepares offline synthesis capability in advance as a backup mechanism. When network connection fails or timing requirements cannot be met, the pre-prepared offline synthesis method provides a reliable fallback, cushioning against connection dependency failures

Inventive Principle:
Principle #11Beforehand cushioning (Prior cushioning)

Data Source

PatentEP4071752B1Text-to-voice processing method and terminal
Publication Date: 2025.08.27 HUAWEI TECH CO LTD
  • EP4071752B1 patent drawingFigure 1~2
  • EP4071752B1 patent drawingFigure 3
  • EP4071752B1 patent drawingFigure 4

AI summary

A text to speech processing method, a terminal, and a server are provided to reduce a speech playing delay of the terminal. The method and the apparatus in this specification may be applied to fields such as artificial intelligence and human-computer interaction. The terminal detects an instruction used to indicate to perform text to speech conversion; sends text to a server; downloads, from the server, audio data converted from the text; determines whether a first frame of playable audio data is downloaded within preset duration; and if the first frame of playable audio data is downloaded within the preset duration, continues to download remaining audio data.