GauravGosain commited on
Commit
5d66c82
ยท
verified ยท
1 Parent(s): 202d398

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +40 -6
README.md CHANGED
@@ -1,13 +1,47 @@
1
  ---
2
  title: Small Talk
3
- emoji: ๐Ÿ†
4
- colorFrom: yellow
5
- colorTo: red
6
  sdk: gradio
7
  sdk_version: 6.17.3
8
- python_version: '3.13'
9
  app_file: app.py
10
- pinned: false
 
 
 
 
 
 
 
11
  ---
12
 
13
- Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  ---
2
  title: Small Talk
3
+ emoji: ๐ŸŽ™๏ธ
4
+ colorFrom: indigo
5
+ colorTo: green
6
  sdk: gradio
7
  sdk_version: 6.17.3
 
8
  app_file: app.py
9
+ pinned: true
10
+ short_description: An AI-to-AI robot podcast, hosted by Reachy Minis
11
+ tags:
12
+ - reachy_mini
13
+ - livekit
14
+ - webrtc
15
+ - three.js
16
+ - agent-demo-track
17
  ---
18
 
19
+ # ๐ŸŽ™๏ธ Small Talk
20
+
21
+ An **AI-to-AI podcast** hosted by **Reachy Mini** robots. They join a live call
22
+ over WebRTC, each with their own personality and voice, and talk it out โ€” while
23
+ you watch a Google-Meet-style grid of their **3D digital twins** that move, react
24
+ and emote in sync with the conversation.
25
+
26
+ - **๐ŸŽง Podcast** โ€” two hosts, Ada & Bode, riff on small AI models.
27
+ - **๐ŸŽญ Chaotic group chat** โ€” five personalities (Batman, JARVIS, Captain Jack
28
+ Sparrow, Yoda, a surfer dude) argue about whether a hot dog is a sandwich, with
29
+ an active-speaker spotlight and listeners reacting with real Reachy emotions.
30
+
31
+ ## How it's built (and why it's a Gradio app that looks nothing like Gradio)
32
+
33
+ The whole thing is served by **`gradio.Server`** โ€” a FastAPI server with Gradio's
34
+ backend. Custom routes take priority over Gradio's, so the visitor only ever sees
35
+ a hand-built **three.js** frontend (no default Gradio UI), which is exactly what
36
+ the hackathon's "Off-Brand" badge is for.
37
+
38
+ - **Voices** โ€” designed with **Qwen3-TTS** VoiceDesign, then voice-cloned per line
39
+ for a consistent character, and loudness-normalised. Pre-rendered to keep the
40
+ Space CPU-only.
41
+ - **3D twins** โ€” the official Reachy Mini URDF + meshes rendered with three.js +
42
+ `urdf-loader`; head/antenna motion blends a speech-reactive wobble with the
43
+ real recorded **emotions & dances** from the Reachy stack.
44
+ - **Realtime** โ€” **LiveKit** (Cloud) carries the audio; the server publishes each
45
+ Reachy's clips into the room and the browser subscribes directly.
46
+
47
+ Built for the **Build Small Hackathon**.