ifmanzhang/huangquan-audiotexttransfer-benchmark
Huangquan AudioTextTransfer Benchmark Artifacts This dataset repository contains preliminary speech-model evaluation artifacts for the Huangquan Road project. It is not the current main game runtime. It records early STT/TTS testing and low-latency speech interaction experiments. Contents README.md: original AudioTextTransfer project notes. docs/tts_model_evaluation.md: model comparison and failure diagnosis. app/, web/, scripts/: local FastAPI + browser test… See the full description on the dataset page: https://huggingface.co/datasets/ifmanzhang/huangquan-audiotexttransfer-benchmark.
Huangquan AudioTextTransfer Benchmark Artifacts
This dataset repository contains preliminary speech-model evaluation artifacts for the Huangquan Road project.
It is not the current main game runtime. It records early STT/TTS testing and low-latency speech interaction experiments.
Contents
README.md: original AudioTextTransfer project notes.docs/tts_model_evaluation.md: model comparison and failure diagnosis.app/,web/,scripts/: local FastAPI + browser test bench source code.runtime/tts/: generated TTS audio samples from testing.runtime/reference/: reference audio used during local tests, where present.runtime/outputs/: generated or intermediate local outputs, where present.
Summary
The prototype tested multiple local STT/TTS options, including SenseVoice/FunASR, Kokoro, VITS/MeloTTS, Piper, Chatterbox-Turbo, VoxCPM2, CosyVoice2/3, Qwen3-TTS, and faster-qwen3-tts.
The frontend includes simulated streaming text output, sentence-level chunking, generation queues, prebuffered playback, first-audio latency measurement, and per-segment generation metrics.
Boundary
Third-party downloaded model weights are not re-hosted here. This repository stores the project's benchmark code, notes, generated audio artifacts, and configuration evidence. Upstream third-party models should be cited separately.