Hot Podcast ·
Muse Finds Its Voice, and a Skill Rolls the Dice
Meta rolls out real-time voice and expressive avatars in its Muse app, and we explore hash-design, a newly discovered skill that uses random hashes to set design directions, as a separate discovery.
AI-generated episode: the script is written with AI from the cited sources and read by synthetic voices; the hosts are fictional.
Prefer your browser’s audio controls? They work here without the global player.

In this episode
Meta Brings Real-Time Voice and Expressive Avatars to Muse
hash-design
qybaihe/hash-design / skills/hash-design
4 repository stars · Data snapshot:
The conversation
AI-generated episode: the script is written with AI from the cited sources and read by synthetic voices; the hosts are fictional.
- Sveva
- Welcome to Hot Podcast! Today an AI assistant learns to talk back in real time, and we also meet a brand new skill with a very unusual way of choosing design. Stefano, what's the news?
- Stefano
- According to Unite.AI, published November 5, 2026, Meta is rolling out real-time voice conversations and expressive avatars in its Muse app in the US, on mobile, web and Mac. Users can interrupt, and the agent keeps working on tasks in the background.
- Sveva
- Interrupting an AI mid-sentence. Finally, a conversation where I can be the rude one!
- Matteo
- Ha! And the report says you can stay voice-only, put the avatar beside the conversation, or go fullscreen. Voice is also described as coming soon to AI Glasses and Muse Charm.
- Stefano
- Now, my analysis, not reported fact: speech plus an avatar could shift assistants from text boxes toward something more like presence. That makes interface design matter more, since the screen must handle listening, talking and background work.
- Sveva
- Any caution there?
- Stefano
- Yes. This is a US rollout, and we only know what the coverage says. We shouldn't assume how well it works in everyday use until people actually test it.
- Matteo
- Now a separate discovery: hash-design, a new skill from a repository created November 1. It has four stars on GitHub as of the November 6 snapshot, which is a repository metric, not a Hot vote. I'm not claiming it works with Muse.
- Stefano
- Per its description, it's for pages, decks, app screens, dashboards, posters and emails that shouldn't look AI-generated. It rolls true-random hashes through an MCP server, freezes several design briefs, then waits for your pick.
- Sveva
- So the dice decide the style, not the model. Brave.
- Matteo
- Purely hypothetical: a designer sketching a mobile screen for a voice assistant could ask for five directions, compare the one-line options, and choose one to build and self-critique.
- Stefano
- At least the randomness is honest. No purple gradient can claim it was my idea.
- Matteo
- Takeaway: voice interfaces raise the design bar, and it's worth exploring tools that push you past generic defaults. Check the details yourself first.
- Sveva
- Thanks, both! You can find this episode's sources on HotSkills. Until next time, keep listening, and keep interrupting politely.