Make Codex a voice-first planning and research experience

I rely on voice as my primary interface. Codex is powerful for implementation, but planning still feels like dictation. I’d like to propose making voice a first-class planning experience with seamless handoff into implementation.
Foundation: make voice first-class. The current Codex mic flow feels record then convert, which introduces pauses that break my flow. I’d love to see continued investment in faster turnaround, smoother turn-taking, and a more conversational feel. That benefits every voice workflow.
Voice first planning workspace : Use the existing sidebar chat as a planning space. Support full-duplex (similar to ChatGPT app on phone) conversation for brainstorming and architecture discussions, while keeping the implementation thread clean. At the end, Codex can propose an artifact, a plan, or a summary, which I can edit and promote into implementation. Dynamic conversational controls. Instead of settings pages, let me adjust in natural language. Be a quiet listener. Challenge my assumptions. Switch to project-aware mode. Controls should be conversational and live, not static.
Long-term vision: Planning should feel like working with a thoughtful engineering partner, while implementation stays focused and structured. The goal is low-friction thinking, clean execution
Thanks
Niranjan

Is this the best location to post for feature requests for Codex? if there is an alternate location, please let me know, I will post it there. Thanks in advance

Codex Product Team, in case if you need further details please reach out to me.

P.S: Currently I use other tools to covert my voice to text to improve the speed of voice to text and then give that out put as input to codex

Thanks for sharing such a thoughtful request, @nbravulapalli. The distinction between voice dictation and a truly conversational planning workspace is especially clear, and the implementation handoff idea is compelling.

We’re sending your request to the team for logging, including your current voice-to-text workaround and the broader vision for voice-first planning.

-Mark G.