Improve Codex CLI input UX: multiline editor-style composer and voice-to-prompt workflow

The main usability bottleneck in Codex CLI is no longer the terminal window itself, but the input field.

I tested terminal-side improvements such as WezTerm and zellij. They work technically, but they do not materially improve the actual prompt-writing experience. The core issue is prompt composition inside Codex CLI.

What is missing:

• comfortable soft-enter / multiline writing
• better editing of long prompts before sending
• draft-style prompt composition
• markdown-first writing experience
• smoother image / clipboard insertion
• a more editor-like interaction model instead of raw shell-style input

What would help:

• soft-enter for newline, separate send shortcut
• embedded multiline markdown composer
• lightweight editor-like UX inside CLI/TUI
• draft mode before sending
• ability to revise larger prompts comfortably
• support for inserting clipboard images or temporary image attachments before send

A useful direction would be a small markdown editor embedded into Codex CLI/TUI. Not a full IDE, but a real composition surface for human-AI collaboration.

The next level would be voice-to-prompt support.

A good reference is whis:
https://github.com/frankdierolf/whis

It supports terminal-oriented voice-to-text flows such as:

whis --as ai-prompt --print

That kind of pattern would be very valuable in Codex CLI:

• speak
• transcribe
• optionally post-process into an AI prompt
• insert into the Codex composer
• review/edit before sending

That would improve both:

• editor-like prompt composition
• speech-to-text input for terminal AI workflows

In short: the biggest UX gain is likely not more terminal tooling, but a much better input/composer model inside Codex CLI itself.