Aispanvok
Voice dictation,
for humans and AI agents
Just talk to your computer. Hold a hotkey anywhere, speak, release — the text lands in whatever field is focused. And because Aispanvok does voice cloning, any AI agent can talk back in a voice you own. A free, fully local alternative to WisprFlow.
macOS 
Windows 
Linux The Captures tab
Every capture, paired with audio and transcript.
Hold the hotkey, talk, release — that's a capture. Replay the original audio, re-transcribe with another model, refine with a local LLM, copy to your clipboard, or send it straight to any MCP-capable agent. Nothing leaves your machine.
Whisper, sized for every machine
Base, Small, Medium, Large, and Turbo — 99 languages, all local, all downloadable in-app.
LLM refinement that respects your words
A local model cleans up filler words, self-corrections, and punctuation — it never rephrases. Raw and refined kept side by side, original audio forever.
Archived by default
Every dictation saves audio and transcript together. Searchable, re-runnable, and one click from becoming a voice-clone sample.
Every agent gets a voice.
One tool call — voicebox.speak — and any MCP-capable agent (Claude Code,
Cursor, Cline…) speaks with a voice you cloned.
{
"mcpServers": {
"aispanvok": {
"url": "http://127.0.0.1:17493/mcp"
}
}
}await aispanvok.speak({
text: "Deploy complete.",
profile: "Morgan"
})
Not on MCP? The same surface is exposed as POST /speak — works with ACP, A2A,
shell scripts, and custom frameworks.
Per-agent voice
Bind a different voice to each MCP client — Claude Code as Morgan, Cursor as Scarlett. You always know who is talking without looking.
Always visible
Every agent-initiated utterance pops a toast. No silent background TTS, ever.
Open protocols
MCP on day one. ACP, A2A, and anything built on tool-call primitives can hit the same endpoint.
Install VoxLoom — Free AI Voice Studio, start dictating.
Free for a limited time, runs locally. No account, no API keys, no per-character fees.