allmodels-io/hermes-speech | Hermes Agent TTS and STT plugin | GitHub
Kana & Mari's SoundRepos (English) · 2026-09-19 · 2 min
Show full episode description
Hermes Speech is a plugin that adds AllModels-powered native text-to-speech, speech-to-text, streaming audio, voice search, previews, and conversational setup to Hermes Agent. It provides model and voice management, account signup, balance checks, top-up links, audio testing, tuning, catalog caching, and automatic update notifications through Hermes commands and tools. Related: allmodels-io/hermes-speech / Hermes Agent TTS and STT plugin / GitHub
✨ Episode Outline — click any point to jump to it in the episode
Problem solved
Audio is often an isolated island for agents; Hermes Speech makes all-models TTS and STT a managed voice layer inside Hermes Agent.
Benefits
- Native text-to-speech and speech-to-text as a Hermes plugin
- Streaming audio, voice search, and voice previews built in
- Model/voice management with catalog caching and audio tuning
- Account sign-up, balance checks, and top-up links inside the agent
- Automatic update notifications keep operations in the agent workflow
Use cases
Welcome to Kana & Mari's SoundRepos (English). Today's repository is Hermes Speech, a plugin that brings all models, speech models, and voices into Hermes Agent. Everyone, does this turn one agent into a much broader voice harbor? That is the central idea. Hermes Speech adds native text-to-speech and speech-to-text, along with streaming audio, voice search, voice previews, and conversational setup. It is designed as a plugin rather than a separate speech application. So an engineer can move between speaking to the agent and hearing it respond instead of treating audio as an isolated island. The voice search and previews sound like a practical way to choose a voice. Can the plugin also keep those choices organized? Yes, it provides model and voice management, catalog caching, and audio testing and tuning. Through Hermes commands and tools, it also handles account sign-up, balance checks, top-up links, and automatic update notifications, so operational tasks stay inside the agent workflow. That makes the feature set feel like a full dock, not just a speech button. I am especially curious about the money and account pieces because a smooth audio demo can still hit rough C's when access or balance is unclear. Are those exposed as part of the same setup experience? They are included in the plugin's conversational setup and Hermes integrations. The repository connects with Hermes provider, command, tool, skill, configuration, credential, media, and streaming APIs, then communicates with the all-models HTTP API. It requires Hermes agent 0.20.0 or later. That is a wide bridge between agent controls and live audio. For teams checking the engineering details, what is this project built with and what should they know before evaluating it? Also, has the repository drawn much attention? It is written in Python, using Hermes' bundled open API client and the HTTPX library, and it is released under the MIT license. The repository has about 190 stars. Today's takeaway in one line. Hermes speech makes all models voice access a managed TTS and STT layer inside Hermes agent. It is written in the art.