Verdict: Voicetypr fits developers who want local-first prompt dictation across apps or machine-readable transcription from a CLI. It is not an agent runtime and it does not submit prompts or control tools for you.
Two documented modes
Mode 1 — focused field
Voicetypr transcribes speech and inserts text into the field that has focus. Its public repository lists Cursor, Claude Code, and ChatGPT as examples. This is input automation, not an agent plugin, and this update did not retest those fields.
Mode 2 — command line
The public v2 source documents status, models, transcribe, and record commands. File and microphone transcription can return a structured artifact for a script to inspect before it decides what to do next.
Developer workflows
Use focused-field insertion for interactive prompts. Use the CLI when a script needs an explicit transcription result it can inspect.
Dictate a longer prompt, inspect the inserted text, then submit it yourself.
Use focused-field dictation for interactive input where normal text paste is accepted.
Call the transcribe command with an explicit file path and request JSON for downstream parsing.
Use the record command's currently documented stop-on-silence mode, then consume plain text or JSON.
Query status or models before choosing whether a workflow can run with the installed local setup.
Route a CLI transcription to an opted-in Voicetypr network-sharing host when that boundary is acceptable.
Transcription CLI
Install the command from Settings → Advanced, then check voicetypr --help on the installed version. Add --json when a script needs structured output.
$ voicetypr transcribe --file note.wav --json
{
"text": "Refactor the auth middleware to use the new token store.",
"words": [ /* structured word data */ ],
"metadata": { /* transcription metadata */ },
"model": "<selected-model>",
"engine": "<selected-engine>"
}
$ voicetypr record --until-silence --json
$ voicetypr transcribe --file clip.wav \
--model whisper --engine local \
--server 192.168.1.20:47842 --password-stdin --jsonKnow the boundary
Last verified: July 27, 2026
This update used source inspection, not a hands-on release test. We reviewed the public CLI parser, launcher, network implementation, README, and release notes. We did not install a macOS or Windows build, execute the CLI, test microphone capture, or connect it to an agent framework. The product-level commands shown here are therefore source-verified; the APP verification gate remains open for environment compatibility and end-to-end automation.
FAQ
No. In interactive use, Voicetypr inserts transcribed text into the focused field. That makes it portable across apps, but it does not give Voicetypr repository context, agent state, or tool access.
A script can invoke the installed Voicetypr CLI. The source documents status, models, transcribe, and record commands, with --json available for machine-readable output. The calling script remains responsible for validation, routing, and any agent action that follows.
The public repository includes versioned HTTP routes used for network transcription, but this review found no separate public compatibility or stability document for direct consumers. Prefer the documented CLI unless you are prepared to track the source-level HTTP implementation.
Local Whisper or Parakeet transcription runs on the device after the selected model is downloaded. Optional cloud speech-to-text and optional AI formatting are different modes with different data flows.
Add --json to a CLI call. For example, voicetypr transcribe --file note.wav --json returns { text, words, metadata, model, engine }. You can also set --model and --engine per call, and use --server <host:port> with an optional --password to route to a running instance.
No. It supplies transcribed text. It does not submit the prompt, choose tools, approve terminal actions, execute an agent loop, or speak the agent's response.
Not for this July 2026 update. We reviewed the public Rust source and release notes, but did not install a release build or execute the commands. Run voicetypr --help and test non-sensitive audio in your own environment before adding it to automation.
Related guides
Narrow to Cursor, inspect the trusted-LAN boundary, or step back to the broader developer dictation workflow.
single-tool focus
For developers evaluating focused-field dictation for Cursor prompts, PR notes, and context.
broader workflow
The general pay-once dictation story for builders deciding whether voice belongs in their daily setup.
trusted-LAN option
Review the network boundary, opt-in sharing behavior, authentication, and source-backed limits before using --server.
broader workflow
Prompts, issues, PR notes, and documentation beyond the automation surface.
Use the 3-day free trial and voicetypr --help to verify your installed release, models, permissions, and output contract.