1 views collected around this technical thread.
This tutorial explains how to create a Python-based speech synthesis tool by installing required packages, configuring Xunfei Open Platform credentials, implementing a Tkinter GUI, and using WebSocket communication to convert text into audio with selectable voice profiles.
GPT‑4o’s speech multimodal system discretizes audio into semantic and acoustic tokens, integrates these tokens with large language models through multi‑stage instruction tuning, and employs hierarchical zero‑shot text‑to‑speech decoding, enabling low‑latency, streaming, and prompt‑driven voice synthesis for applications like gaming.
This tutorial walks through setting up Spring AI 0.8.1 with Spring Boot 3.1.1, configuring Alibaba Tongyi model access, and implementing chat, streaming, image, and audio generation endpoints using Java code and vector database integrations.
This tutorial demonstrates how to obtain Baidu TTS credentials, construct the required signature, send an HTTP request using PHP's cURL library, and save the returned audio data as an MP3 file, providing a complete code example for developers.
This article introduces the AI sales assistant from 58.com, detailing its background, a few‑shot voice cloning approach using real dialogue data, multi‑accent naturalness optimization, deployment architecture, and future plans, while evaluating performance metrics and discussing challenges in speech synthesis quality and stability.
The article presents 58 Tongcheng AI Lab's AI sales assistant, detailing its background, a few‑shot voice‑cloning pipeline built on real dialogue data, data preprocessing, FastSpeech2‑based acoustic modeling, multi‑accent style transfer, deployment architecture, controllable synthesis parameters, and future research directions.
This step‑by‑step tutorial shows how to set up a Stable Diffusion web UI, install the required Python environment and GPU‑enabled PyTorch, add Chinese localization and optional LoRA or Deforum extensions, generate realistic human images, create animated videos, and add speech with D‑ID, all ready for deployment on your own AI website.
This article provides a comprehensive overview of Volcano Engine's virtual digital human platform, detailing its definition, AI‑driven and human‑driven classifications, 2D and 3D technical architectures, multi‑modal perception, interaction capabilities, application scenarios, and future development directions.
iQIYI’s IQDubbing system leverages AI‑driven voice conversion to automatically generate high‑quality, expressive dubbing in dozens of languages and over 50 character voice styles, streamlining multilingual film and TV localization, reducing reliance on scarce actors, and earning positive audience feedback, patents and industry awards.
The MSMC‑TTS system, a multi‑stage multi‑codebook VQ‑VAE based neural text‑to‑speech solution, delivers near‑human audio quality (MOS 4.41) with a compact 3.12 MB acoustic model, substantially surpassing Mel‑Spectrogram FastSpeech baselines in naturalness and efficiency.
This article reviews Alibaba's digital‑human (virtual avatar) research over the past few years, covering the product’s evolution, a six‑stage pipeline for building digital humans, solutions to key challenges in realism, multimodal interaction, and the open‑source MMTK algorithm library.
This tutorial explains how to create a lightweight speech synthesis tool using iFlytek's WebAPI, covering required environment setup, API credential acquisition, GUI design with Tkinter, and detailed Python code for WebSocket communication, audio handling, and WAV file generation.
This article provides a hands‑on guide to using the pyttsx3 library for offline text‑to‑speech conversion in Python, covering installation, basic playback, voice property adjustments, multilingual support, and conditional speech examples with counters.
The iQIYI M2VoC competition at ICASSP 2021, the first low‑resource multi‑speaker, multi‑style voice‑cloning challenge, attracted 153 academic and industry teams to tackle few‑shot (100 utterances) and extreme few‑shot (5 utterances) tracks, evaluated by professional listeners, yielding strong innovations and applications while confirming that single‑sample cloning remains unsolved.
The paper presents a speaker‑aware module (SAM) that enables high‑quality voice conversion using only a single utterance of the target speaker, addressing the small‑data challenge in speech timbre transfer and achieving state‑of‑the‑art performance on the Aishell‑1 benchmark.
The iQIYI M2VoC Challenge at ICASSP 2021 invites researchers to tackle low‑resource multi‑speaker, multi‑style voice cloning by providing Mandarin datasets, few‑shot and extremely few‑shot tracks with strict data rules, MOS‑based subjective evaluation, and a $9,600 prize pool for top submissions.
The article reviews Interspeech 2019, summarizing major trends and representative papers in end‑to‑end speech recognition, synthesis, natural language understanding, speaker recognition, and speech translation, while also highlighting best student papers and providing resources for further study.
This article provides a comprehensive overview of voice conversion technology, covering its definition, parallel and non‑parallel data approaches, classic and deep‑learning methods such as DTW, GMM, seq2seq, PPG, VAE, Flow, GAN, and practical applications and challenges in iQIYI’s products.
Voice conversion (VC) technology changes a speaker’s timbre and style while keeping the spoken text unchanged, supporting one‑to‑one, many‑to‑one, and many‑to‑many scenarios for medical assistance and entertainment, using parallel or non‑parallel data through methods such as DTW‑aligned frame mapping, attention‑based neural networks, PPG‑LSTM pipelines, VAEs, normalizing‑flow models, and GANs, with iQIYI focusing on non‑parallel data, prosody preservation, and noise‑robust augmentation.
This article reviews recent advances in low‑resource text‑to‑speech synthesis, covering the background of TTS, challenges in data‑ and compute‑limited scenarios, and detailed descriptions of FastSpeech, LightTTS, LightBERT, and related lightweight vocoder techniques, along with experimental results and future research directions.