The Problem
Cursor IDE has changed how developers write code by putting AI collaboration at the center of the development workflow. But there is a bottleneck that most Cursor users encounter quickly: the AI is only as good as the instructions you give it, and writing detailed instructions is slow.
When you open Cursor's AI chat or use the inline edit feature, you are essentially writing a mini specification every time you want the AI to do something meaningful. "Refactor this function to use async/await, handle the error cases explicitly, add JSDoc comments, and make sure it follows the pattern used in the auth module." That instruction takes time to type. And because typing is slow, most developers write shorter, less precise instructions — and get less precise code back.
The same applies to code documentation. Cursor can generate documentation, but to get documentation that truly captures the intent, context, and constraints of your code, you need to describe what you know. Typing that description breaks the flow of development.
How Telvr Works with Cursor
Telvr works with Cursor IDE through system-level text injection, with no Cursor plugin or extension required. The integration is seamless because Telvr operates at the OS level — it inserts text at whatever field has focus, whether that is the Cursor AI chat, the inline edit prompt, a code comment, or a markdown documentation file.
Here are the three primary workflows in Cursor:
AI Chat prompting: Click into the Cursor AI chat sidebar, press your Telvr hotkey, and speak your development request in detail. The text appears in the chat field, ready to send.
Inline edit (Cmd+K): Trigger Cursor's inline edit with Cmd+K, then use Telvr to speak your edit instruction. The instruction appears in the inline prompt field, and Cursor processes it with its AI.
Code documentation: Position your cursor above a function or class in the editor, press the Telvr hotkey, and dictate your documentation block directly into the code.
All three workflows use the same hotkey and the same Telvr interface — you just need your cursor in the right field before pressing the key.
Best Enrichment Mode for Cursor
Dev Task mode is the most powerful choice for Cursor AI chat interactions. When you speak a development request and process it through Dev Task, the output is structured around the four elements that AI coding assistants respond to best: context, requirements, implementation notes, and acceptance criteria.
A structured prompt in Cursor's AI chat produces significantly better code than a vague instruction. Dev Task mode structures your spoken request into precisely the format that makes AI code generation most effective.
Raw Transcription is the right choice for Cursor's inline edit prompts (Cmd+K). Inline edits are typically precise, targeted instructions: "Extract this logic into a separate function," or "Add null checks for all parameters." These instructions do not need restructuring — they need to arrive in the AI's context exactly as you intended. Raw mode gives you a clean transcription of your spoken instruction without adding structure that the inline edit flow does not need.
Use Dev Task for longer, more architectural requests in the AI chat. Use Raw for quick, precise inline edits.
Before and After Example
Raw speech input (Dev Task mode — Cursor AI chat):
"I need you to build a custom React hook for managing form state uh it should handle validation with a schema that I can pass in as a prop and it needs to support async validation for things like checking if a username is already taken um the hook should return the field values, errors, a submit handler, and also a loading state for the async validation and it should work with TypeScript generics so the field types are inferred"
After Dev Task mode:
Context Building a custom React hook for form state management with schema-based validation support. Needs to handle both synchronous and asynchronous validation cases (e.g., server-side uniqueness checks).
Requirements
- Accept a validation schema as a parameter (define appropriate schema interface)
- Support async validation callbacks for individual fields
- Return: field values, field errors, submit handler, async validation loading state
- Full TypeScript generic support so field types are inferred from the schema
Implementation Notes
- Async validation should be debounced to avoid excessive requests
- Loading state should be per-field, not global, for granular UI feedback
Acceptance Criteria
- Hook compiles cleanly with strict TypeScript
- Sync and async validators both surface errors in the errors return value
- Loading state correctly reflects in-flight async validation
- Submit handler prevents submission when validation errors are present
Cursor's AI will produce substantially more accurate and complete code from this structured prompt than from the raw spoken version.
Time Savings
Cursor users interact with AI constantly throughout their workday. Conservative estimates put AI chat interactions at twenty to forty per day for active Cursor users. Each interaction involves writing an instruction.
The difference between a thirty-second spoken prompt processed by Telvr and a three-minute typed prompt is significant at that volume. Twenty interactions per day at two and a half minutes saved each equals fifty minutes recovered — every day.
Beyond raw time, the quality improvement compounds. Better-structured prompts produce better initial code, which means fewer iterations to get to a working result. Each iteration you save through better first-attempt prompts eliminates multiple rounds of back-and-forth with the AI.
Cursor is built on the premise that AI collaboration should be frictionless. Telvr removes the last major friction point in that workflow: the time and effort required to communicate clearly with the AI.
Getting Started
- Download Telvr for macOS from telvr.ai and complete setup.
- Configure your microphone and set a push-to-talk hotkey — something that does not conflict with Cursor's existing shortcuts.
- Open Cursor and start a new AI chat session.
- Set Dev Task as your default mode for chat prompts, Raw for inline edits.
- Try your first voice prompt: describe a component or function you need and see the structured output in the chat field.
The combination of Cursor and Telvr creates a development workflow where you spend more time thinking and less time typing. Ideas go from your head to working code faster than any keyboard-based workflow allows.
New users get a €3 Welcome Credit to try all modes with no commitment. After that, you only pay for what you use — tiered pricing starts at €0.030 per minute and drops to as low as €0.003 per minute as your usage grows. No monthly minimum — a small cost relative to the development time it recovers.