Conomo

Conomo is a compact streaming speech-recognition package for local, on-device transcription. It is distributed by Memo for use with Memo Agent.

Download

• Conomo ASR INT8 v1 archive — approximately 373 MB to download and 411 MB extracted.
• Checksums and file manifest.

Memo Agent downloads and verifies this package during setup. No Hugging Face account or token is needed to download the public files. Speech recognition runs locally after installation.

Package

The archive contains ONNX speech-recognition components, tokenizer/configuration files, a speech detector, and license/provenance notices. It accepts 16 kHz mono audio and supports streaming partial transcription. Weight storage uses a compact mixed-precision representation with additional INT8 storage conversion; computation remains FP32 in the supported runtime.

This is a model package, not a hosted transcription endpoint. It requires a compatible speech runtime. Local compatibility and small-corpus checks do not establish accuracy across every speaker, language or acoustic environment.

Privacy

The package contains no user recordings, transcripts, credentials or personal configuration. The weights were not fine-tuned on Memo users’ recordings or transcription history.

License

Model materials are distributed under OpenMDW-1.1. The included speech detector has its own MIT license. Required notices of origin and attribution are retained inside the archive. See the bundled license and notice files.