SDK Overview
Hecttor ships two speech-enhancement SDKs that solve different problems for different listeners. Pick the one that matches who (or what) is on the receiving end of your audio.
Two SDKs, two listeners
Hermes for humansVoice & video callsPodcastsVoice interfacesHearing assist
Use it when a person is on the other end of the audio pipeline. Combines noise cancellation, speech-speed adjustment, and voice boost; each is independent and toggleable on its own.
Orpheus for machinesASR pipelinesVoice AI agentsReal-time transcriptionWER reduction
Use it when audio is feeding an ASR (automatic speech recognition) model, a voice agent, or any downstream model that consumes speech. Tuned for WER improvement, not for subjective listening quality.
Both SDKs are real-time, chunk-based, and run locally inside your application. No audio is streamed to Hecttor for processing.
Choosing between Hermes and Orpheus
| Your use case | SDK |
|---|---|
| Voice / video conferencing | Hermes |
| Voicemail / podcast / recording cleanup | Hermes |
| Hearing-assist features | Hermes |
| Pre-ASR audio cleanup (cloud or on-device ASR) | Orpheus |
| Voice-AI agents (Pipecat, LiveKit Agents, etc.) | Orpheus |
| Real-time transcription pipelines | Orpheus |
If you need both (for example, a meeting product that records audio for humans and transcribes it), you can run Hermes and Orpheus side by side on the same input stream.
What's in this section
- Licensing: how SDK access works and what an evaluation license includes.
- Security and Deployment: where processing happens, what the SDK needs from the host environment, and how to think about data handling.
- Evaluations: how we recommend measuring the impact of each SDK on your pipeline.