Skip to content
SPECIFICTION
Get listed

Confucius4-TTS

github.com

WebLinuxAPISelf-hosted
Visit github.com
Spec score 7.9 #119 of 233 in Text-to-speech tools

Spec sheet

Purpose
Confucius4-TTS is an LLM-based text-to-speech system for multilingual and cross-lingual speech synthesis that preserves speaker identity across languages.Read 13 Oct 2026 · github.com
Product
Confucius4-TTS is an LLM-based multilingual and cross-lingual zero-shot text-to-speech engine.Read 12 Oct 2026 · github.com
Platforms
web,api,linux,self_hostedRead 20 Sep 2026 · github.com
Deployment
The repository provides a Gradio web UI and a FastAPI server with streaming and non-streaming endpoints.Read 12 Oct 2026 · github.com
API
A FastAPI server provides non-streaming and streaming TTS endpoints that accept text, a language code and a reference audio upload.Read 13 Oct 2026 · github.com
Integrations
The project acknowledges use of Qwen3-TTS, CosyVoice, Amphion/MaskGCT, w2v-BERT 2.0, Seed-VC, and BigVGAN components or methods.Read 12 Oct 2026 · github.com
License
The repository includes an Apache License, Version 2.0.Read 13 Oct 2026 · github.com
Maker
The GitHub organization identifies itself as NetEase Youdao and lists Beijing as its location.Read 13 Oct 2026 · github.com
API access
YesRead 20 Sep 2026 · github.com
API output
The non-streaming API returns a WAV file and the streaming API returns raw int16 little-endian PCM chunks.Read 12 Oct 2026 · github.com
Commercial use
YesRead 20 Sep 2026 · github.com
Cross-language transfer
The project says users can synthesize speech across its supported languages while retaining the same voice.Read 13 Oct 2026 · github.com
Cross-lingual transfer
It can synthesize speech across its supported languages while retaining speaker identity.Read 12 Oct 2026 · github.com
Emotion
The README describes emotion transfer as a feature.Read 12 Oct 2026 · github.com
Emotion transfer
The listed features include emotion transfer that aims to carry the feeling of a reference voice into generated speech.Read 13 Oct 2026 · github.com
Export formats
WAV,PCMRead 20 Sep 2026 · github.com
Fine-tuning
The README documents training workflows for its Text2Semantic and Semantic2Acoustic modules using prepared TSV training data and pretrained models.Read 13 Oct 2026 · github.com
Languages
14Read 20 Sep 2026 · github.com
Languages
The project lists support for 14 languages: Chinese, English, Japanese, Korean, German, French, Spanish, Indonesian, Italian, Thai, Portuguese, Russian, Malay and Vietnamese.Read 13 Oct 2026 · github.com
Limit
The README says vLLM support currently targets version 0.16.0 with the V1 engine and warns that hardware and CUDA requirements may prevent it from running.Read 12 Oct 2026 · github.com
Performance
The README reports multilingual and cross-lingual benchmark results, including WER, CER, and speaker similarity metrics.Read 12 Oct 2026 · github.com
Requirements
The listed installation requirements are Python 3.10 and CUDA 12.6.Read 13 Oct 2026 · github.com
Streaming
The project documents streaming generation that returns raw int16 little-endian PCM chunks and provides the sample rate in a response header.Read 13 Oct 2026 · github.com
vLLM support
For faster Text2Semantic generation, the README documents a vLLM backend and says it currently supports vLLM 0.16.0 with the V1 engine.Read 13 Oct 2026 · github.com
Voice cloning
It supports zero-shot voice cloning without requiring a reference transcript or additional training.Read 13 Oct 2026 · github.com
Web demo
The README links to an online Gradio demo and describes a browser UI where users upload reference audio, enter text, select a language and generate speech.Read 13 Oct 2026 · github.com

Where Specifiction places it

Alternatives to Confucius4-TTSAll alternatives →