Voice Recognition Tuned for Developers

Context-tuned recognition for common software and IT vocabulary

Enable the Software Development profile to bias speech recognition toward common developer tools, protocols, and infrastructure terms.

Download AICHE
Works on
macOSWindowsLinuxiOSAndroid

Short answer: enable the Software Development profile in your client settings (Pro tier) to supply common developer terminology directly to the transcription engine, keeping tools, protocols, and cloud nouns accurate without manual spelling edits.

The problem this solves

Standard voice models are trained on conversational prose. When you speak engineering concepts, generic models frequently transcribe technical nouns as phonetically similar everyday words: hearing "Kubernetes" as "cooper net ease", "OAuth" as "o off", or "gRPC" as "G RPC".

The Software Development profile (it_engineering) provides domain-specific vocabulary context to both the speech transcription pass and the subsequent spelling-correction prompt.

How it works

  1. Available on the Pro tier.
  2. Select Software Development in your client app's settings.
  3. Your client app stores this preference locally and passes the profile: "it_engineering" parameter with each audio transcription request.
  4. When you record audio (such as with the global hotkey ⌃+⌥+R on macOS or Ctrl+Alt+R on Windows and Linux), the backend combines your custom vocabulary with the curated profile terms.
  5. The combined vocabulary cue guides Whisper transcription and the spelling-correction stage, helping recognize specialized tools and acronyms accurately.

Curated technical vocabulary

The backend includes a fixed set of curated software engineering terms and phonetic hints:

  • AI and developer tools: Claude Code, Codex, ChatGPT, Anthropic, LLM, Whisper, GitHub
  • Protocols and authentication: OAuth, JWT, GraphQL, gRPC
  • Infrastructure and cloud: Kubernetes, nginx, Docker, CI/CD, AWS, Terraform
  • Databases and messaging: PostgreSQL, Redis, MongoDB, Kafka
  • Languages and ecosystems: React, Swift, Kotlin, npm
  • Engineering operations: Jira, Figma, Sentry, Datadog

Priority and Custom Vocabulary pairing

Custom vocabulary and the software profile work together within a 50-entry recognition window:

  1. Custom vocabulary comes first: Any words you define in your personal dictionary take precedence.
  2. Profile terms fill remaining slots: Curated software terms populate the remaining available slots up to the 50-entry limit.
  3. Deduplication: Terms overlapping between your personal dictionary and the profile are automatically deduplicated.

This means you can reserve custom vocabulary slots for your specific internal microservices, private repository names, and team members, while the Software Development profile covers common industry technologies.

Scope and technical boundaries

To keep expectations clear, here is what the Software Development profile is and is not designed to do:

  • Best-effort context cue: The profile provides recognition priors and spelling hints. It improves accuracy for common technical terms, but cannot guarantee recognition of every arbitrary phrase.
  • Audio transcription only: The profile applies to audio transcription paths. Standalone text enhancement does not currently evaluate this profile.
  • Not a syntax formatter: The profile does not act as an AST parser, code generator, or case rewriter. It does not automatically transform words into camelCase or snake_case, format CLI strings, or validate programming language grammar.
  • Client-persisted preference: The selected profile is maintained locally by the client application sending the audio request; it is not stored as an account-level setting on the backend database.

Tips for best results

  • Add unique stack names to Custom Vocabulary: While common tools like Docker and Redis are built into the profile, unique project names, custom package names, or brand-new libraries belong in your personal dictionary.
  • Speak naturally: Dictate thoughts, bug explanations, and prompt descriptions as normal prose. The model uses the vocabulary context to resolve ambiguous technical pronunciations.
  • Use Voice Code for coding assistants: Pair developer vocabulary with AICHE's Voice Code mode when prompting AI coding tools like Claude Code, Codex, or Cursor for smooth, hands-free agent workflows.

Works With

The Software Development profile enhances voice input wherever you dictate developer workflows:

  • Claude Code - technical context in AI coding prompts
  • Cursor - developer tool names in architecture instructions
  • VS Code - documentation with tool names intact
  • Terminal - commit messages and explanations
  • GitHub - PR descriptions with proper technology nouns

Tags

ai-codingdevelopmentideproductivity

Try AICHE Free

Stop typing. Start speaking.

Global hotkey dictation with zero audio retention. Free 7-day trial, no credit card required.