Instead of typing to Copilot, try talking to it (nicely). Click the microphone icon in the GitHub Copilot app to transcribe your prompt. 🎙️

Download the app ⬇️

"(nicely)" How most people will probably talk with it: "COPILOT HOW DID YOU FUCK UP THIS BADLY YOU ABSOLUTE MORON MATE? IT WAS **ONE LINE OF CODE** AND YOU FUCKED IT UP OH MY GOD IT IS BAFFLING HOW DUMB YOU ARE OH MY FUCKING GOD WHAT IS WRONG WITH YOU"

Great explanation! 🎙️ Thanks for breaking down the new voice feature so clearly — and yes, I’ll make sure to talk *nicely* to Copilot, as requested 😂 You made it sound so smooth and natural. Looking forward to more updates like this! 🌊🧊

SAAR SAAAAAR do not REDEEM

What if I don’t play nice 😑

Voice input for prompts is a small change that adds up over a day.

Honestly, I still feel weird talking out loud to my IDE, but the transcription is getting way too good to ignore.🤩

Voice is great for explaining intent and terrible for file paths, symbols, and exact constraints. The winning interface will let me talk through the problem, then edit the transcript before Copilot confidently mishears one variable name.

It will go mad if I talk to it when it messes the things even after specific instructions 3 times :)

Voice is a faster input channel, not a safer authority channel. Before Copilot runs a consequential command, the screen should restate the repo, branch, files and intended side effect in text. Speech can propose the action; a visual receipt should still authorize it.

Can't do that sitting in the office

Jesus you fucking clowns, 3% of the people who actually pay for that dumpster fire of Chinese crap and Indian Jeets who lied .. cannot wait for something else to replace you

I think the sharper point is this. Thinking about GitHub's transcription changes the frame for me. Voice becomes an interface when it can carry intent, not just sound.

Don't tell us what to do or how to do it.

I think the sharper point is this. Thinking about GitHub's transcription changes the frame for me. Voice becomes an interface when it can carry intent, not just sound.

Your support is trash 🚮

This would mean to me that I’d be talking 6h a day, I did the in the past sitting in meetings, no thanks

Excellent work so far. I'd love to hear more and discuss opportunities 🔥

Voice removes typing friction and adds interpretation risk. For any command that edits files, opens a pull request, or runs a tool, Copilot should display the exact transcript and planned action before execution. A misheard sentence is funny in chat and expensive in a repository.

I tried voice prompts in my IDE once and my cat walked by, yelled into the mic, and Copilot refactored my auth middleware into a loop.

Voice prompts make coding feel much more natural. Excited to try this!

Voice input fits this really well—some ideas are much easier to explain than type. The transcription quality will make or break it, though.

Voice may become the fastest way to express intent. The real gain comes when Copilot turns an ambiguous request into a plan the developer can inspect before execution.

...and give Microsoft ALSO your voice imprint 💁‍♂️ Then just an image and a clone of you is created! 🫣

Please confirm that the co-pilot can run $Doom now.

Voice makes prompting faster, but a transcript is still an instruction with consequences. For anything that edits or publishes, I would show the exact interpreted command and risky actions before execution. Speech errors are harmless in chat and expensive in a repository. GitHub Copilot can keep that confirmation lightweight.

voice is faster until noisy offices turn variable names into silent bugs

I like the approach here. The groundwork is coming together. DM me if you'd like to talk promotion.

Can you add an animated pfp function?