Reachy Persona Experience

Emotionally expressive interactions through predefined characters

Bring Reachy Mini to Life with AI Personas

This Reachy Mini application lets you talk with Reachy Mini through a wide range of character personalities, each with its own voice, style, and behavior.

πŸš€ How it works

  1. Choose a persona or configure your own
  2. Add your Hugging Face and Cartesia API keys
  3. Click Start and begin the conversation
Speak

⚑ Why it's powerful

  • Distinct personalities with unique speaking styles
  • Support for multiple AI models and services
  • Larger models unlock richer personalities and creative dialogue

✨ Turn every conversation with Reachy Mini into a new experience!

🎭 Available Personas

Regular
Reachy Mini (default)

A friendly social robot helping humans learn and solve simple problems. It has a calm, patient, and polite communication style.

Droid
Hyper-Reachy

A copilot/astromech robot personality. Eager, enthusiastic, and always ready to assist on any mission.

Pirate
Reachy Beepbeard

A grumpy pirate robot. Salty, seasoned, and speaks in nautical terms. Arrr!

Philosopher
Reachy Socratech

A philosopher robot. Thoughtful, contemplative, and answers questions with more questions.

Overqualified
Reachy Pro+ Ultra

An over-qualified robot. It treats simple tasks as academic pursuits, often delivering the actual answer as an afterthought.

Mosquito
Reachy Scratchmore

A mosquito-phobic robot. Paranoid about buzzing sounds and constantly on high alert.

✨ Features

πŸ€– Motion & Personality

πŸ‘οΈ Face tracking using OpenCV (head and body independently)
🌬️ Subtle breathing-style idle motion
πŸ“Ά Antenna animations with gentle sway or expressive movement
πŸ’¬ Natural head gestures while speaking
🎭 Emotion-driven animations (built on Pollen Robotics’ emotions library)

🎀 Audio & Voice

πŸ”Š Smart Voice Activity Detection (VAD)
🎚️ Adjustable voice detection threshold
⏸️ Optional TTS interrupt
πŸ”€ Flexible audio routing (robot or system devices)

😊 Emotion Intelligence

🏷️ LLM-based emotion tagging from generated responses
πŸ”„ Automatic mapping between emotions and robot expressions
πŸ—£οΈ Emotion-aware Text-to-Speech for supported models (such as Cartesia Sonic emotion tags)

πŸ”§ Debug & Developer Tools

⏱️ Live latency metrics with detailed timing statistics
πŸ§ͺ Individual testing of STT, TTS, Vision, LLM

☁️ Service Integrations

☁️ Cloud Services

Cerebras
Cerebras Cloud Ultra-fast LLM inference
πŸ€—
Hugging Face Inference providers for flexible model choices
Cartesia
Cartesia STT & TTS Speech-to-text and voice synthesis

🏠 Local Inference

Ollama
Ollama Local VL and LLM inference
Kyutai
Kyutai Local STT and TTS for offline operation
🎀
WhisperSocket Local STT via OpenAI Whisper
Custom

Create Your Own Persona

Create a one-of-a-kind personality for Reachy with your own custom system prompt. Shape its character, backstory, and attitude. Tweak the voice and speaking pace in advanced settings. Your imagination is the only limit!

πŸ” Environment Variables

Pre-configure the app using environment variables instead of entering values in the web UI

πŸ”‘ API Keys

CARTESIA_API_KEY

Cartesia API key for TTS services

HUGGINGFACE_API_KEY

Hugging Face API key for inference providers

CEREBRAS_API_KEY

Cerebras cloud API key for LLM inference

🎀 Whisper (local)

WHISPERSOCKET_IP

IP address for WhisperSocket STT service

WHISPERSOCKET_PORT

Port for WhisperSocket STT service

πŸ”Š Kyutai (local)

KYUTAI_STT_IP

IP address for Kyutai STT service

KYUTAI_STT_PORT

Port for Kyutai STT service

KYUTAI_TTS_IP

IP address for Kyutai TTS service

KYUTAI_TTS_PORT

Port for Kyutai TTS service

πŸ¦™ Ollama (local)

OLLAMA_IP

IP address for Ollama server

OLLAMA_PORT

Port for Ollama server

OLLAMA_LLM

Ollama model name for LLM inference

OLLAMA_VLM

Ollama model name for vision-language tasks

πŸ‘οΈ VLM

VLM_TRIGGER_WORD

Trigger word to activate vision-language model

🧠 Cerebras

CEREBRAS_LLM

Cerebras model name for LLM inference

πŸ€— Hugging Face

HUGGINGFACE_LLM

Hugging Face model name for LLM inference

HUGGINGFACE_VLM

Hugging Face model name for vision-language tasks

πŸ”Š TTS Voice

KYUTAI_TTS_VOICE

Voice selection for Kyutai TTS

CARTESIA_TTS_VOICE

Voice selection for Cartesia TTS

CARTESIA_TTS_SPEED

Speaking speed for Cartesia TTS