How do I switch from local (3GB) to cloud model?

[AFROAgencyAI]

Hey guys been using FluidVoice and am very happy with it but it takes 5-10s before the text appears in a chatbox. Might be because my M2 MacBook can't handle the local processing. How do I switch models? I can't find the option either in the app or online

[Etogerman]

FluidVoice has two separate model settings, which makes this a little confusing:

• Voice Engine controls speech-to-text. These models run locally.
• AI Enhancement is optional post-processing. It can use local Fluid Intelligence or a cloud provider such as OpenAI or Groq.

If the model you downloaded is Whisper Large (~2.9 GB), that likely explains the delay. The large Whisper models are non-streaming, so final transcription happens only after you stop speaking.

Open FluidVoice → Voice Engine, then select, download, and activate a faster model:

• For English: Parakeet Flash (~250 MB) or Parakeet TDT v2 (~500 MB).
• For multilingual dictation: Parakeet TDT v3 (~500 MB) or Nemotron Speech 3.5 (~670 MB).
• You can also try the built-in Apple Speech model, which requires no download.

If the 3 GB download was actually Fluid Intelligence (~3.5 GB), open AI Enhancement instead. You can disable local enhancement or configure OpenAI/Groq with an API key. That switches only the transcript enhancement to the cloud; speech recognition itself remains local.

A useful check: which model is currently shown under Voice Engine → Active Model?

References: supported models in v1.6.6 and streaming support logic.