🎯 Starting budget: 100,000 UAH
It is necessary to develop an AI assistant (likely based on the RAG model), which will respond to the user and send responses to the AI Avatar (via D-ID API).
The avatar, in turn, will produce a video response (video rendering is fully performed by the D-ID API).
📘 D-ID API documentation: https://docs.d-id.com/reference/get-started
1) Frontend
A simple interface needs to be developed, which will display 5 roles (video avatars).
Each avatar is connected and managed through the D-ID API.
🔹 Front example: https://roliki.ua/s/script/neil/neilavatar-main.html
2) Backend (CMS)
A control panel where the following are stored:
all main API settings (keys, parameters, etc.);
tools for training and configuring the assistant.
3) Training the AI assistant (RAG model)
a) Fundamental training (manual)
At the development stage, we set the assistant's basic knowledge — about our hero, his roles, communication style, and rules.
This is done once, without the need for a separate tool.
b) Automatic training from the blog
The assistant automatically receives new articles from WordPress (REST API) — title + text.
The system itself converts them into vector data and adds them to the knowledge base, updating the avatar's intelligence without manual intervention.
c) Manual training through CMS
In the admin panel, texts and knowledge for the assistant can be added manually.
d) General knowledge about the world
For natural responses, the assistant must also possess general knowledge.
Approximately: 80% — project knowledge base, 20% — general data (ChatGPT).
This makes the responses lively and contextual.
💡 Two options are possible:
own RAG model,
or using D-ID Agent API.
Advantages of working with me
✅ Detailed feedback-recommendation
📲 Always in touch (Mon–Fri, 10:00–20:00)
🔍 Deep immersion in the project
⏱ Adequate attitude towards deadlines
💸 Bonus 10–20% for quality implementation
