Chatbot to generate text and image

• 

  Example Conversation

Inspiration

I thought of this because in the past ChatGPT was not able to generate image you want straight in the app and have to use plugins.

What it does

This basically functions like weaker version of ChatGPT but it has the ability to generate image.

How we built it

This project was built on models from HuggingFace by using their inference API

Challenges we ran into

The model in HuggingFace is weaker compare to GPT3.5/4 unless I was using Llama from Facebook. But I had get permission to use it and I did not had enough time to request permission.

Accomplishments that we're proud of

I'm proud that I can do something like this so quick with Streamlit. Building Chatbots and data related stuff with Streamlit is indeed very easy and developer friendly.

What we learned

I learned to use part of Langchain to connect with HuggingFace. Nowadays the AI trend is becoming the norms. Normal people would have to get on the train before it's too late.

What's next for Chatbot to generate text and image

Maybe including speech-to-text would be nice, but considering OpenAI is going to achieve all these sooner or later or actually at the time of writing these they have already released all the features I was doing including this speec-to-text. But we couldn't get free version, building with HuggingFace allows people to experience for free and if they were to think of going all out they can consider paying to ChatGPT. All in all a lot of features in ChatGPT is still in beta so we still need experts to properly prompt the bot so we can get the information we wants.

Built With

• api
• git
• github
• huggingface
• python
• streamlit
• streamlit-community-cloud