First the audio is transcribed and the voice of who is speaking is detected. It is then sent to a Large Language Model that gives a response, the text is then translated (optional step if the language isn’t English) and lastly a text-to-speech model outputs the conversation. It’s important to note that this is a conversation: the previous text is always joined (or concactenated) with the previous reponses and questions. The prototype is designed to have a memory so it knows what is being said. Speech-to-Text : The AI captures the user’s spoken words and translates them into written text, using advanced algorithms to accurately interpret various languages, accents, and speech nuances. Natural Language Processing (NLP) : NLP technology analyzes the converted text, understanding the context, sentiment, and intention behind the words. It draws from extensive linguistic data to comprehend grammar, idioms, and colloquial expressions. Language Model : This is where our AI generates a response. It uses a language model, such as GPT-4 , to formulate an appropriate and contextually relevant response. Text-to-Speech : The written response is then converted back into spoken words, using synthetic voices that are increasingly indistinguishable from human speech.

Talking Pictures

Revolutionize Art

Interactive, AI-Powered Conversations.

We make artworks speak for themselves.

Because a picture says more than a 1000 words.

Let art speak to you

How visitors respond

Click here for the video with Dutch subtitles.

This article on the MSK project in one of the biggest Belgian News agency (article in Dutch) details Talking Pictures

Let’s talk!

If our project resonates with you and you see potential for a collaboration, we would 💙 to hear from you.

EDUCATE

INTERACT

ENTERTAIN

Imagine your gallery where each artwork is like a guide, ready to chat, joke, and share their tales with every visitor, making art fun, relatable, and super interactive!

With ‘Talking Pictures’, you’re not just hanging art; you’re creating lively, chatty companions that turn every visit into a memorable, engaging adventure, ensuring visitors leave with a smile, a story, and a heap of new knowledge tucked into their experience.

Interested?

We’re looking for partners!

Are you a curator? Do you own a gallery, or work for one? Would you like to see Talking Pictures in a particular venue? Would you like to collaborate?

Tell us what you need, it would be our pleasure to provide more info and think of the best solutions.

Promo Video

Video Demonstration

During development, we have focussed on art, but the applications are endless. Talking Pictures can be an asset in almost any sector, including education, HR, healthcare, social work…

We are currently developing more models, so stay tuned or reach out so we can tell you all about them.

FAQ

How do the paintings talk?

simple non technical diagram of conversational AI with audio input - conversation -audio output

Through something called conversational AI, which is a subfield of artificial intelligence that enables machines to understand, process, and respond to human language in a natural and intuitive manner. Using sophisticated machine learning models, conversational AI can comprehend spoken or written language, process the information, generate appropriate responses, and communicate them back to the user.

Our project takes this a step further, integrating conversational AI into everyday objects, allowing them to engage in meaningful dialogue with users. Imagine a painting that can discuss its history, a toy that can converse with a child, or even a coffee machine that knows your favorite brew.

How does it work?

First the audio is transcribed and the voice of who is speaking is detected. It is then sent to a Large Language Model that gives a response, the text is then translated (optional step if the language isn’t English) and lastly a text-to-speech model outputs the conversation.

It’s important to note that this is a conversation: the previous text is always joined (or concactenated) with the previous reponses and questions. The prototype is designed to have a memory so it knows what is being said.

Speech-to-Text: The AI captures the user’s spoken words and translates them into written text, using advanced algorithms to accurately interpret various languages, accents, and speech nuances.
Natural Language Processing (NLP): NLP technology analyzes the converted text, understanding the context, sentiment, and intention behind the words. It draws from extensive linguistic data to comprehend grammar, idioms, and colloquial expressions.
Language Model: This is where our AI generates a response. It uses a language model, such as GPT-4, to formulate an appropriate and contextually relevant response.
Text-to-Speech: The written response is then converted back into spoken words, using synthetic voices that are increasingly indistinguishable from human speech.

Let’s talk!

If our project resonates with you and you see potential for a collaboration, we would 💙 to hear from you.