README.md

metadata

title: Small Talk
emoji: 🎙️
colorFrom: indigo
colorTo: green
sdk: gradio
sdk_version: 6.17.3
app_file: app.py
pinned: true
short_description: An AI-to-AI robot podcast hosted by Reachy Minis
tags:
  - reachy_mini
  - livekit
  - webrtc
  - three.js
  - track:wood
  - sponsor:nvidia
  - sponsor:modal
  - achievement:offbrand
  - achievement:llama
  - achievement:fieldnotes
  - achievement:offgrid
  - badge-tiny-titan

Small Talk

An AI-to-AI podcast hosted by Reachy Mini robots. They join a live WebRTC call, each with its own personality and voice, and talk it out while you watch a Meet-style grid of their 3D digital twins moving in sync with the conversation. Give them a topic and they write the script, design their own voices, dress themselves, and go live. Own a Reachy Mini? It can join a show as a real cast member and speak its lines through the actual robot.

Team: GauravGosain and nkapila6.

• Demo video: https://youtu.be/obP4C1eH77I
• Build write-ups: Small Talk on the Hugging Face blog · nkapila.me
• Launch posts: @_GauravGosain on X · Nikhil Kapila on LinkedIn

What you can do

• Watch a live generated show. Pick a topic. One structured Nemotron call writes the cast and the full speaker-to-dialogue script, Qwen3-TTS voices each line, and the next line renders while the current one plays. Subtitles, a pre-show "writers' room", and rolling continuations keep it going.
• Set the cast. A slider picks 2 to 5 hosts, or how many simulated co-hosts fill in around your physical robots.
• Design a robot. Choose a name, personality, voice, shell colour, and props. The same Nemotron brain styles its wardrobe from your description.
• Tune into Reachy FM. A radio station of AI-written songs with synced karaoke lyrics, a spinning vinyl deck, an audio-reactive visualizer, and a DJ robot in headphones that does mic breaks and bops to the beat.
• Bring your own Reachy. A single Go binary turns a physical Reachy Mini into a cast member that speaks its own lines, head and antennas moving with the speech.

How it is built

The whole app is served by gradio.Server, a FastAPI host with Gradio's backend where custom routes take priority, so the visitor only ever sees a hand-built three.js frontend. There is no default Gradio component anywhere in the product.

flowchart LR
    topic([Topic]) --> llm["NVIDIA Nemotron 4B<br/>llama.cpp on Modal"]
    llm -->|one structured call| script[["Cast plus script<br/>(JSON)"]]
    script --> tts["Qwen3-TTS<br/>on Modal"]
    tts -->|"line N+1 renders<br/>while line N plays"| pub[ReachyPublisher]
    pub --> sfu{{"LiveKit SFU<br/>(WebRTC)"}}
    sfu --> web["Browser:<br/>3D twins + subtitles"]
    sfu --> robot["Physical Reachy<br/>(Go companion)"]

• Brain. NVIDIA Nemotron Nano (4B) served through llama.cpp on Modal. A single constrained, structured call returns the full cast and script as JSON. We found constrained structured output far more reliable than chaining calls.
• Voice. Qwen3-TTS VoiceDesign on Modal, one consistent character voice per host, generated as a cascade so there is no dead air between lines.
• Realtime. A self-hosted LiveKit SFU carries the audio over WebRTC. Subtitles and show status ride LiveKit data messages.
• Twins. The official Reachy Mini URDF and meshes in three.js, with head and antenna motion blending a speech-reactive envelope and the real recorded Reachy emotions and dances.

The Space itself runs CPU-only. All inference is delegated to Modal serverless GPUs. The Modal serving code for the Nemotron (llama.cpp) and Qwen3-TTS endpoints lives in nkapila6/llama-modal-serve.

Built for the Build Small Hackathon

Everything runs on models well under the 32B cap, and most of the work is done by a single 4B model. Small Talk is in the running for:

Category	Why it qualifies
Thousand Token Wood	A whimsical, AI-native entertainment platform.
NVIDIA	The brain is NVIDIA Nemotron.
Modal	The LLM and the TTS both run on Modal at runtime.
Off Brand	A fully custom three.js UI built on gradio.Server.
Tiny Titan	The reasoning brain is a 4B model.
Llama Champion	Nemotron is served through the llama.cpp runtime.
Off the Grid	No proprietary or closed model APIs. Every model (Nemotron, Qwen3-TTS) is open-weight and self-hosted via llama.cpp; Modal provides the compute, not the model.
Field Notes	A full build write-up is published on the HF blog.
Bonus Quest Champion	The most bonus criteria met across the board.

If you enjoy it, an upvote helps with Community Choice.

Source: github.com/Gaurav-Gosain/small-talk