- Language coverage
- Determined by the model family you install from the signed catalog
- Deployment
- Local Desktop (macOS/Windows) or open-core CLI and local HTTP API
- Privacy default
- No account, no telemetry, no central cloud STT backend
- Offline control
- No silent auto-download on normal STT; CLI --offline fails closed
Chinese audio on hardware you control
Teams handling Mandarin meetings, bilingual interviews, or Chinese content drafts often need ASR that does not ship audio to a third-party region they did not choose. OpenASR keeps recognition local after you install the appropriate catalog models.
Pick model families from the signed catalog that match your language mix; do not assume one weight covers every dialect or domain.
Multilingual work is model-led
OpenASR is infrastructure for local ASR, not a single fixed language pack. Catalog families differ in language support and features such as Voice ID marking in the marketplace.
Install deliberately, verify with your own samples, and use CLI --offline when the environment must not reach the network.
- Install language-capable models from the signed catalog
- Transcribe locally on Desktop or via CLI/API
- Add speaker separation and optional Voice ID with any local file ASR model on Desktop
- Automate with --diarize and /v1/voice-id/* when building pipelines
Desktop vs open core for Chinese workflows
Desktop suits producers and operators on Mac or Windows who want a local app, speaker separation, and named speakers via Voice ID with any local file ASR model.
Researchers and backend owners can standardize on the open-core CLI and local HTTP API across machines, including Linux hosts without a native desktop app.
What we do not claim
These pages do not invent WER numbers, do not crown a universal “best Chinese ASR,” and do not describe a Linux native desktop client. Validate quality on your accent, domain vocabulary, and microphone chain.
Frequently asked questions
Can OpenASR transcribe Chinese fully offline?
Yes, when the Chinese-capable models you need are already installed locally. The stack is local-first; CLI --offline fails closed rather than downloading or calling a cloud STT backend.
Is Chinese support a separate cloud product?
No. Language capability comes from the local model families in the signed catalog, used through Desktop or the open core on your machines.
Does it only support Chinese?
No. Multilingual coverage depends on the catalog model family you choose. Install the families that match your languages; do not assume a single default covers all locales.
Can I separate speakers on a Chinese meeting recording?
Yes. Desktop supports speaker separation and optional Voice ID naming with every local file ASR model. The CLI exposes anonymous --diarize and the local API exposes operator-only /v1/voice-id/* routes for automation.
Is there a Linux desktop app for Chinese ASR?
No native Linux desktop app. Use the Apache-2.0 CLI, local HTTP API, or container images on Linux.
Will models auto-download the first time I paste Chinese audio?
Normal transcription is designed not to silently auto-download models. You install from the signed catalog on purpose, which keeps air-gapped and regulated setups predictable.