TL;DR: Standard Voice Mode may be going away, but the feel doesn’t have to. Say, Pi for ChatGPT brings back the calm, patient voice experience people loved — inside ChatGPT itself. It’s familiar on purpose: same voices, thoughtful pacing, and now with hands-free interruptions. Available today on Edge, Firefox, and Chrome.

Why we built this (and why now)

When people said “Save Standard Voice,” they weren’t asking for flashy prosody — they were asking for presence: full, steady answers in a consistent voice you could trust at 2 a.m. That’s what our MVP restores.

This isn’t a sudden pivot for us. Back in September 2023, when ChatGPT’s Standard Voice Mode first launched, Say, Pi already had a near-identical experience for Pi.ai — in fact, we shipped our version three days earlier. Both approaches were (and are) powered by Whisper for accurate speech recognition, with the same design philosophy: be patient, listen well, and don’t rush people.

Now we’re bringing that same Standard Voice feel to ChatGPT.

What you can expect in the MVP

Familiar where it matters

Same voices, same vibe. We trigger ChatGPT’s own Read Aloud so you hear the voices you already know.
Thoughtful, full answers. This mirrors Standard Voice Mode’s steadier conversational pace.

Better where it helps

Hands-free interruptions (on by default). Start talking naturally — we pause, listen, and adapt without hunting for a button.
Stay in the chat UI. Keep the standard ChatGPT interface in view while you talk (great for following long text, copying code, etc.).
Patient end-pointing. Our speech end-detection is tuned to avoid cutting you off.

Quick start (it’s simple)

Install the Say, Pi extension (Edge, Firefox, Chrome).
Open chatgpt.com and click the Call button.
Talk. We’ll listen, transcribe with Whisper, and trigger Read Aloud for the AI’s reply.
Want to use your voice anywhere? Right-click any text field → Start typing with Say, Pi.

What’s different from Advanced Voice Mode

This isn’t the ultra-low-latency “talk over me” style. It’s the Standard style — deliberate and calm. Many people prefer that: less interruption, more depth, more trust.

Latency expectations: It’s not instant. We favor signal over speed.
Style expectations: Less chirpy, more steady. You’ll hear the voice you chose, delivering the same kind of full answers you used to get.

Honest limitations (and how we’ll evolve)

We want you to know exactly what you’re getting — and why.

Read-Aloud dependency (MVP):
To keep this free (and simple) at launch, we reuse ChatGPT’s Read Aloud in the web app.
- Great for shorter replies (e.g., GPT-5’s brisk answers): audio starts soon after text is ready.
- Slower for long replies (e.g., GPT-4o-style essays): we must wait for the full text to finish streaming before audio generation starts. That can mean tens of seconds (even a minute+) for very long answers.
- Roadmap: We’ll add true audio streaming so speech begins while text is still generating. That introduces real inference cost, so it will likely live in Premium. We’ll be transparent about pricing.
September 9 UI risk:
OpenAI’s changes around the Standard Voice sunset may alter the ChatGPT UI (e.g., Read Aloud placement/behavior). If something breaks, expect brief turbulence.
- Our plan B: If needed, we’ll switch to generating audio ourselves with a TTS pipeline to preserve the same voices and streaming behavior. If that happens, we’ll ship a fix fast — and explain clearly what changed.

How close is it to “old” Standard Voice?

Very close by design.

Same voices you know from ChatGPT’s Read Aloud → same timbre and delivery.
Comparable recognition quality (Whisper).
More patient listening (our end-pointing is tuned to wait that extra beat).
Stay in the standard chat view (no forced full-screen). A dedicated focus mode is planned later for those who want it.

If anything feels “off,” tell us. We built this for the people who loved Standard Voice; your notes guide our polish.

What’s next

Streaming speech while text is still generating (Premium).
Optional focus/full-screen mode on ChatGPT (we already support this for Pi).
Resilience against UI shifts, with a seamless fall-back TTS path.
Improved mobile support for Firefox and Kiwi on Android. For now, you'll find the best experience on desktop.

Install now

For the #SaveStandardVoice community

You proved something important: connection isn’t disposable. Say, Pi for ChatGPT is our way of honoring that — not by arguing on social media, but by shipping the experience you asked for.

Keep the Standard Voice feel in ChatGPT.
We’ll keep listening. ❤️

Whisper is a trademark of its respective owner. ChatGPT is a trademark of OpenAI. Say, Pi is an independent product and is not affiliated with OpenAI.

Say, Pi for ChatGPT — Saving Standard Voice

🎧 Listen to this post