Reachy Stories

Reachy Stories

Lets a child make up a bedtime story with a Reachy Mini robot: the robot asks three questions, the child answers by voice, and the robot writes and reads the story aloud. For families and robot developers.

Made by pollen-robotics on Hugging Face. These notes are Mindket's, written from its Space page and README; the app's own files and license are on Hugging Face.

About

What it is: Reachy Stories lets a child make up a bedtime story together with a Reachy Mini robot. The robot asks three short questions about a hero, a place and a magical object. The child answers by voice, and the robot then writes a full story timed to a target length and reads it aloud. The heavy work runs on a ZeroGPU backend.

Who it suits: It suits families and educators who own a Reachy Mini, and developers who want to study a voice-driven robot app.

Title: Who might use this app
Industry | Who uses it | How they use it | Example
Family life | Parents | Make up a bedtime story with a child | The child names a hero, a place and a magical object
Education | Teachers | Use a robot storyteller for a listening activity | Run one story with a small group
Robotics | Developers | Study a robot app that listens and speaks | Read how the robot link is set up over WebRTC

If you do not have a Reachy Mini robot, you can skip it, since the narration plays through the robot and not the browser.

Where to use it: You use a browser to start it and the robot to hear it.

• Web browser: the Space's page on Hugging Face, where you sign in and pick your robot
• Reachy Mini robot: its microphone hears the child and its speaker plays the story
• Windows, macOS or Linux: a local copy that runs on Python 3.12

Title: What the app is good for
Use case | Who | What they get back
Co-create a bedtime story | Children and parents | A complete story, narrated by the robot
Story from a picture | Families | A story that Gemma writes from a photo the robot camera took
Build on the robot link | Developers | A working example of a robot connected over WebRTC

How to set it up: The hosted Space needs only a robot and an account. A local copy needs more.

• A Reachy Mini robot
• A Hugging Face account, for the sign-in step
• Python 3.12, with pip
• Node and npm, only if you change the browser code under the ts folder

1. Install the huggingface_hub package.
2. Download the Space's files into a folder.
3. Move into that folder.
4. Install Gradio 6.15.2.
5. Install the packages from requirements.txt.
6. Start the app with Python.

pip install -U huggingface_hub
hf download pollen-robotics/reachy-stories --repo-type space --local-dir reachy-stories
cd reachy-stories
pip install gradio==6.15.2
pip install -r requirements.txt
python app.py

If you change the browser code, rebuild it so the prebuilt files in the static folder stay current.

cd ts && npm install && npm run build   # outputs to ../static/

How to use it: Open the Space online, or start your copy with the command above. Sign in with Hugging Face, pick your robot, and let the robot ask its three questions. The child answers out loud, then listens to the story.

Title: Parts of the repository
Tool | What it does
app.py | The backend that opens a session for each robot
central_consumer.py | Receives the robot's microphone audio and sends voice back to its speaker
gpu_worker.py | Runs the GPU work as separate calls
ts | The browser code for sign-in and robot selection, built into the static folder

Example: A child tells the robot that the hero is a brave fox, the place is a snowy forest and the magical object is a lantern. The robot writes a story around those three pieces and reads it aloud through its speaker.

For newcomers and experts: If you are new to this, begin with the three questions and keep the answers short. Everything the child hears comes from the robot, so keep the robot close.

If you know the field, one worker thread per session runs the script in three stages, interview, generate and narrate. GPU jobs run as separate calls: Gemma pulls out each spoken answer and then writes the whole story. Kokoro synthesizes the narration in bounded chunks, so a narration lasting minutes never holds the GPU continuously. The robot wobbles its head during narration and flicks its antennas on a browser message.

The child answers three questions by voice
Gemma extracts each answer
Gemma writes the full story
Kokoro narrates it in chunks
The robot speaks and moves

Good to know: The audio goes to the robot, and the browser only handles sign-in.

• Sound: nothing plays in the browser
• Sign-in: the browser sends your Hugging Face token, the robot's id and a connection offer to the backend
• Camera: Gemma can also write a story from a picture that the robot camera takes
• Voice: Kokoro speaks at 24 kHz
• License: apache-2
• Cost to run: a Reachy Mini robot, a Hugging Face account and ZeroGPU time, and no other paid service is named
• Data it reaches: the child's spoken answers from the robot microphone, any camera picture, and your Hugging Face token
• Not for: playing the story in the browser, because the audio goes to the robot speaker

On Hugging Face, 13 November 2026

• Likes: 1
• Built with: gradio
• License: apache-2.0

Links

• Space: https://huggingface.co/spaces/pollen-robotics/reachy-stories
• Files: https://huggingface.co/spaces/pollen-robotics/reachy-stories/tree/main

Try it

Open the Space in your browser, or copy it to your own Hugging Face account from its page. Nothing in this kit runs by itself.