Voxta v1.6.1: HY-Motion, F5-TTS & MCP Hotfixes

Hello everyone,

A quick hotfix on top of v1.6.0. This one cleans up the rough edges from the v1.6.0 rollout — HY-Motion now ships with its own dependencies (no more Orpheus workaround), the VaM broken-voice race condition is fixed, F5-TTS no longer crashes on libtorchcodec, and MCP gets several robustness fixes.

🕺 HY-Motion

• Ships with its own package dependencies now — fresh installs work without installing Orpheus first

• "Lite" option fixed (was broken in v1.6.0)

🎙️ VaM compatibility

• Fixed race condition causing broken voice in file-based audio clients (most visible in VaM)

• Chat loading messages now emit to file-based audio clients — pairs with the new chat-startup status overlay in voxta-virtamate v2.0.1 (package #98), so VaM users finally get visible progress between Start Chat and the first audio chunk

🔊 F5-TTS

• Reference clip auto-transcription now uses the same approach as Orpheus, fixing "Could not load libtorchcodec"

🧠 MCP

• Stdio MCP servers are now disposed cleanly on shutdown

• MCP child process no longer polluted by AutoRun (uses CMD /D /C), fixing ClientTransportClosedException

• UI warns on duplicate MCP augmentation keys

• Importing an MCP preset now refreshes the items list

🤗 Hugging Face

• HF declarations can now specify a model subfolder using //

🖥️ Voxta UI

• Inspector display improvements

📦 Voxta Installer

• Prevents uninstall while Voxta is running, so you can't end up with a half-broken install

After subscribing to the required Patreon tier, you can get the latest version at: https://portal.voxta.ai/

Thanks for reporting these so quickly after v1.6.0 — keep the feedback coming.