Voice
Confirmed EvoX voice capabilities fall into two groups: composer dictation and the speech-generation plugin for speech synthesis, voice cloning, and audio transcription. This is not real-time two-way voice conversation.
Dictate a prompt
Select the microphone in the composer and speak. The result becomes editable text, so review names, numbers, paths, and punctuation before sending. Dictation requires operating-system microphone permission and a working input device.
Use the speech plugin
The built-in speech-generation plugin starts disabled. Configure a supported provider before enabling it. Remote audio upload, voice cloning, and potentially billable calls retain explicit approval.
| Tool | Purpose |
|---|---|
list_voices | List provider and locally registered voices. |
preview_voice | Create a short preview audio file. |
synthesize_speech | Generate complete speech audio. |
clone_voice | Create a reusable voice from reference and consent recordings. |
forget_voice | Remove a local registration and distinguish any supported remote deletion. |
transcribe_audio | Produce SRT or text; local whisper.cpp can be the default backend. |
Audio outputs go to the task artifact directory and transcripts to the transcript directory. The plugin does not accept an arbitrary output directory.
Voice rights and recording safety
Clone only the user's own voice or a voice with explicit authorization. Provider consent text must be genuinely spoken by the user; EvoX cannot generate or accept the legal declaration on their behalf.
Local transcription does not upload the recording. Choosing a remote provider should identify the destination in the approval card. The plugin does not provide live voice conversion, calls, or real-time voice chat.
EvoX Docs · Features · Capabilities