Give your coding tools more context. Type less of it.

Why voice works well for technical prompts

Developers often know what needs to be explained but compress the prompt because typing the full context feels slow. Voice changes that cost. It is easier to describe the current behaviour, expected behaviour, constraints, files involved, and edge cases in one spoken pass.

The goal is not to replace precise technical writing. It is to capture the complete thought quickly, then turn it into structured text. A good dictated prompt can contain the same information you would give a teammate during a careful handoff.

A practical developer dictation workflow

1. Choose Raw for exact capture or Polished for a cleaner technical explanation.
2. Hold the shortcut and describe the task in complete sections, including constraints.
3. Release, review the copied text, and paste it into your coding assistant or issue.

For sensitive code or client context, the speech recognition remains local. If you use a built-in profile, its text processing can also run locally. The destination tool may still receive whatever text you choose to paste, so its own privacy and retention rules continue to apply.

Use vocabulary for frameworks, tools, and product names

Technical speech contains terms that ordinary language models may interpret as common words. Elokin ships with a development vocabulary covering languages, frameworks, infrastructure tools, databases, AI tooling, and common platform names.

You can extend it with the technologies used in your own stack: internal service names, package names, acronyms, repositories, and product terminology. Import a text file with one term per line to add an entire glossary at once. Choose Merge to retain existing entries without duplicates, or Replace when the imported file should become the whole category.

Separate categories make large lists easier to maintain. You can clear one category, clear all vocabulary, or restore the default dictionaries from Settings.

Dictate more than AI prompts

• Bug reports: explain reproduction steps and expected behaviour while the sequence is fresh.
• Pull request descriptions: summarise the motivation, implementation, risk, and testing.
• Documentation: draft an explanation in conversational language before refining it.
• Code review notes: capture reasoning without interrupting the review to type a long paragraph.
• Stand-up updates: turn a spoken recap into a concise written status.
• Architecture notes: record trade-offs and rejected options while exploring a design.

Local models first, custom AI when useful

Elokin’s default profiles are local because transcription and routine cleanup do not always justify an external request. This avoids spending provider tokens just to convert speech or fix punctuation.

For workflows that need more reasoning, you can create custom profiles backed by OpenAI, Anthropic, or an OpenAI-compatible endpoint. For example, a custom profile could turn a dictated bug description into a specific issue template. That cloud processing is optional: you decide which provider, model, prompt, and profile should use it.

This separation keeps the common path predictable: local speech recognition first, external intelligence only when the transformation genuinely benefits from it.

Frequently asked questions