Skip to main content
Enconvo uses separate speech-model settings for different jobs. Choose the recognition model in the feature you are actually using.

​
Choose a model

Open the corresponding model settings, choose a provider, and review language and model options. Cloud Plan variants use your Enconvo account; direct providers use their own configured access. Local MLX and CoreML options process recognition on-device after required models are available. See the complete speech-provider catalog for the declared providers, including cloud services and local Parakeet, Qwen, Whisper, SenseVoice, Paraformer, and Nemotron variants. Hardware and availability depend on the selected runtime and build.

​
Match capabilities to the task

  • Streaming: produces partial text during recording; use a realtime-capable model.
  • Batch transcription: processes a file after capture/upload.
  • Language hints and hot words: guide recognition when the provider supports them.
  • Speaker diarization: separates voices; it does not guarantee the correct person’s name.
  • Live translation: only offered by supported providers/models, with source and target options.
  • Timestamps and structured output: depend on the provider and transcription output format.

​
Verify with a small sample

Test a short recording with the language, microphone, and terms you actually use. Review the result before processing a long meeting. For a failure, read the error for file-size, format, extraction, account, or network details. Model downloads can be much larger than the app. Check the actual download size and free space instead of assuming every local speech model has the same requirements. Related: Speech Assets, Permissions, Local Models.