Edge TTS
A tool that calls the online neural text-to-speech service built into the Microsoft Edge browser. It requires no local model or GPU and offers low latency and broad language coverage, but provides only preset voices and does not support voice cloning.
Timeline (last 90 days)
IboChat's /speak endpoint tries three voice paths in order: GPT-SoVITS (best voice cloning, takes about 5–18 seconds, supports only English, Chinese, Japanese, and Korean, and permanently occupies about 3.9GB of VRAM once loaded), OmniVoice Studio (supports any language, takes about 30–40 seconds, called via an OpenAI-compatible endpoint), and Edge TTS (Microsoft's free neural voices, no cloning capability, supports 24 languages, zero latency, never fails).
Edge TTS refers to calling the online neural speech synthesis service built into Microsoft's Edge browser (often accessed via unofficial client libraries such as edge-tts); it requires no local model or GPU, offers very low latency and broad language coverage, but only provides preset voices and cannot clone a user-specified voice.
When using a cheap model (such as the glm-4.5-air level) and replying only with text to each message, the LLM cost of the conversation is just a rounding error; Edge TTS is free, and the two GPU models consume electricity.
All Facts (3)
IboChat's /speak endpoint tries three voice paths in order: GPT-SoVITS (best voice cloning, takes about 5–18 seconds, supports only English, Chinese, Japanese, and Korean, and permanently occupies about 3.9GB of VRAM once loaded), OmniVoice Studio (supports any language, takes about 30–40 seconds, called via an OpenAI-compatible endpoint), and Edge TTS (Microsoft's free neural voices, no cloning capability, supports 24 languages, zero latency, never fails).
50%UnverifiedEdge TTS refers to calling the online neural speech synthesis service built into Microsoft's Edge browser (often accessed via unofficial client libraries such as edge-tts); it requires no local model or GPU, offers very low latency and broad language coverage, but only provides preset voices and cannot clone a user-specified voice.
50%UnverifiedWhen using a cheap model (such as the glm-4.5-air level) and replying only with text to each message, the LLM cost of the conversation is just a rounding error; Edge TTS is free, and the two GPU models consume electricity.
50%