iTranslated by AI
Improving Efficiency with Voice Input Using Monologue
The iOS version of the voice input app Monologue has been released. As someone who has been utilizing the Mac version to give instructions to AI agents, I will introduce the expanded use cases made possible by the iOS version and the convenience of a custom mode for Claude Code.
Target Audience
- Engineers who use AI agents (Claude Code, Cursor, etc.) on a daily basis
- Those interested in voice input but hesitant due to concerns about accuracy
- Those who want to record ideas during commutes or in their spare time
Background and Purpose
In development using AI agents, the quality of code is directly influenced by the instructions in natural language. Typing long instructions on a keyboard can be unexpectedly tedious, and it's not uncommon to settle for vague instructions with a "this will do" attitude.
The voice input app Monologue, provided by Every, solved this problem. Its characteristic feature is not just simple voice recognition, but its ability to "organize spoken content into proper text."
I have been a fan of the Mac version for a while, but with the release of the iOS version in February 2026, the scope of its application has expanded significantly.
What is Monologue?
Monologue is an app that converts voice into "polished text." While standard Apple voice input "transcribes spoken words as they are," Monologue translates the spoken content into "written language."
Key Features
- Context-Aware Formatting — Automatically adjusts to be casual for iMessage, structured for email, or bulleted for Notes.
- Whisper Mode — Recognizable even in a whisper. Usable in cafes or offices.
- Personal Dictionary Learning — Remembers friends' names, technical terms, project names, etc.
- Sync with Mac Version — Vocabulary data is shared between Mac and iOS.
Since the iOS version operates as a keyboard extension, it can be used directly within any app, such as iMessage, Gmail, Slack, and Notes.
Using the Mac Version for AI Agent Instructions
The Mac version of Monologue has been particularly effective for giving instructions to Claude Code.
Facing the terminal and speaking, "Write a test for this component. Coverage should be... let's see... in a way that covers major branches, oh, and include error cases," Monologue organizes it as follows:
Please write tests for this component. Ensure major branches are covered and include error cases.
Even if you stumble, rephrase, or say "uhm," it corrects it into proper Japanese. This is truly helpful.
With keyboard input, you tend to brace yourself, thinking "I have to write concisely," but with voice, you can naturally give instructions following your flow of thought. As a result, the information volume of the instructions increased, and the output quality of the AI agent also improved.
Expanded Use Cases with the iOS Version
With the release of the iOS version, Monologue can now be used even when you're away from your PC.
Recording Ideas and Reminders
I’m now recording ideas that pop into my head while walking or on the way to buy coffee, using voice input on the spot. This is more comfortable than I imagined.
Previously, I would use flick input on a memo app, but typing while walking was tedious, and I often ended up forgetting things thinking "I'll write it down later." With Monologue, I can just speak while walking and it leaves behind well-organized text.
"That API error handling, uh, maybe we should make the number of retries configurable. Also, the timeout value shouldn't be hard-coded; it should read from the config file."
Even if I speak ramblingly like this, it gets recorded as a clean reminder. The key point is that when I get back to my desk, I can use that content directly as an instruction for Claude Code.
Task Organization During Commutes
Organizing today's to-do list by speaking aloud during my commute was also effective. Dumping your thoughts through voice helps organize your thinking and makes priorities clearer.
Logical Instructions with Claude Code Mode
Monologue has a feature to create custom modes tailored to specific purposes. Setting up a "Claude Code Mode" here makes it even more convenient.
While the normal mode just "organizes spoken content into natural sentences," the Claude Code mode converts it into a logical instruction sequence.
For example, suppose I said this:
"Add an avatar image upload feature to the user profile page. Oh, server-side image resizing might be better too. And validation is needed, let's limit the file size to 5MB. Right, and make it so it uploads to S3."
Through Claude Code mode, it is converted as follows:
Please add an avatar image upload feature to the user profile page.
Requirements:
- Upload destination: S3
- File size limit: 5MB (implement validation)
- Resize images on the server-side
Even if the order of speaking is jumbled, it organizes it into a logical structure that is easy for the AI agent to process. Even if infrastructure, validation, and UI topics are mixed together, it groups them by category.
This allows for high-quality prompts just by speaking in the order thoughts occur, without having to be conscious of the "quality of instructions."
Challenges and Tips
Summary
- With the release of Monologue iOS, you can now create polished text using voice input even when you're not in front of your PC.
- It's ideal for giving instructions to AI agents because it organizes spoken content into natural Japanese even if you stumble or rephrase.
- The habit of recording quick ideas or reminders by voice improves idea management for development.
- Setting up a Claude Code mode converts your thoughts into a logical instruction structure, even if you speak them in a jumbled order.
- The flow of "Speak → Refine → Instruct" is becoming a new interface in the era of AI agents.
Voice input isn't just a "substitute for when your hands are full," but the most natural way to output your thoughts. Now that collaborating with AI agents has become a standard, I feel that tools like Monologue will significantly change the development experience.
Discussion