Talk, Tell & Automate: Introducing Continuous Speech‐to‐Prompt
You will never write to an AI again.
Samuel ROY & Olivier Legris
September 19, 2025
Hey everyone,
This week we’re launching a brand new way to interact with Alter: Continuous Speech‑to‑Prompt.
Before, using Speech‑to‑Prompt did automatically start a new Alter conversation window.
Not anymore. You can now interact with an Alter conversation window only by speaking.
This opens a brand new world of possibilities where your Alter experience is more conversational and seamless than ever.
How It Works
Speak freely – Just hit and hold the “mic” button (or use your global shortcut) and start talking. Release when finished.
Auto‑inject – Your words are turned into a prompt and dropped straight into Alter’s active conversation window.
Instant execution – The command is sent immediately, so you get a fast reply.
If no Alter conversation is open, a new conversation opens automatically. If one is already active, your voice prompt lands there, no extra clicks, no context switching.
What You Can Do
Scenario | Example |
Orchestrate workflows | “Schedule a meeting with the team for tomorrow at 3 PM.” “Now send an iMessage to Sam to let him know about the meeting.” |
Ask questions back and forth | “What’s the latest on project X?” “Condense it in 1 paragraphs “Now create a sum up email about it” |
All voice instructions can be transcribed locally (using Whisper Pro or Parakeet) for extra speed and privacy.
Why It Matters
Hands‑free productivity – Keep your hands on the keyboard for typing, but let your voice handle routine commands.
Seamless context switching – No need to open a new tab; Alter knows where you’re talking.
Fast local inference – Zero latency between speaking and seeing results.
🦸 The super hero trick : Organise a movie screening
Browse the internet, then tell Alter to help you choose the movie, create the calendar invite, and send iMessages about the screening.
🙏 We Want Your Feedback
How can we make this feature more useful to you? Let us know by replying here or on our Discord
Cheers
Full Changelog
New Features & Enhancements
Continuous speech-to-prompt: Keep chatting by voice with your last conversation window
Conversation Window Resizing: Resize conversation windows both in width and height
Conversation Performance: Dramatically improved conversation loading with smooth scrolling
YouTube Integration: Faster YouTube transcript fetching
Voice Triggers: Trigger actions by voice using keywords or short sentences
Notch Interface: Added convenient login/logout button directly in the notch header
Meeting Preferences: Keep-awake functionality is now a user preference in Voice settings
Speaker Recognition: Filter Speakers for faster identification and manual renaming
Bug Fixes & Stability
Audio System: Fixed microphone muting when system audio is active
Model Selection: The "best" model now properly reselects at every conversation turn
Notch Display: Fixed disappearing notches on notifications and improved screen configuration handling
X-Callback URLs: Prevented settings windows from opening on alter:// URLs
Cloud Transcription: Better error display and handling for transcription failures
Conversation Management: Enhanced input focus behavior and send button visibility
Crash fixes: Multiple stability fixes including conversation search, and context handling
Speech-to-Prompt: Acting on text selections with speech-to-prompt is back!
Audio Processing: Refactored audio processor with improved tap audio processing as default
Localization: Updated French translations
Dependencies: Updated OpenAI library to avoid JSON parsing failures