Project Overview
Talk Insight is an AI-powered communication analysis service that helps users better understand
their conversations through emotional tone detection, attachment style analysis, and conflict
trigger identification. Users upload a chat file — either exported text or a screenshot — and Talk
Insight returns a detailed report on the communication dynamics. It also suggests healthier
response strategies and offers an AI chatbot to rehearse more thoughtful replies. The service is
especially useful for conversations that feel emotionally heavy or unclear, such as those with
family members, partners, or colleagues. Talk Insight aims to build a society where everyone can
communicate more clearly, empathetically, and safely.
Talk to Insight, at talk-insight.com!
Identifying the Challenge
The Social Problem:
Many people today experience emotional stress and misunderstandings in digital
communication, particularly with close relationships. Miscommunication often leads to conflicts,
emotional burnout, or even long-term mental health issues. In Asia, where indirect
communication is common and emotional expression is restrained, this issue is even more
significant.
A 2019 study in Korea showed that when families resolve conflicts with verbal aggression, the
risk of depression increases by 58%. For women, the risk goes up even more — by 2.3 times.
What if we could catch those patterns earlier?
Talk Insight addresses this communication gap by providing insight into users’ messaging
patterns and how they can improve — helping prevent conflict before it escalates.
Innovation and Uniqueness
Why Our Project Stands Out:
Unlike typical emotion analysis tools, Talk Insight provides a multilayered approach. It not only
analyzes emotional tones but also identifies attachment styles and toxic patterns such as blame or
avoidance. It supports both individual and group chats, screenshots and text files, and multiple
languages. Talk Insight also includes an AI chatbot that suggests questions and responses based
on past conversations, and the entire system adjusts to the user's tone and device preferences —
including dark/light mode sync with system settings. It’s not just analysis — it’s a step toward
healthier communication habits.
Insights and Development
Learning Journey:
At first, we focused on emotion detection alone, but user feedback helped us realize the value of
analyzing attachment and conflict styles. We also learned how much users appreciated AIsuggested replies during emotionally difficult conversations. One insight was that users often
struggle to identify themselves in chat data, which led us to build a “Who are you?” selector into
the onboarding flow. Small design changes like this had a big impact.
We also built on top of OpenAI’s API to focus on usability and real-world problem solving. One
major insight came from early user testing: too many features at once left users confused. So we
redesigned the interface to guide users naturally, creating an “auto-onboarding” flow where the
design itself explains how the tool works. This pivot taught us that great UX isn’t just about
aesthetics—it’s about removing friction and making users feel supported from the very first click.
Development Process:
Talk Insight is an AI-powered communication analysis tool designed to help users navigate
interpersonal challenges and enhance relationships. It analyzes conversations to detect emotional
tone, toxic language, and conflict patterns, then offers actionable feedback to foster healthier
communication.
We built Talk Insight as a full-stack PWA using Next.js, React, and Postgres, hosted on Vercel,
with a strong emphasis on usability, approachability, and security. Users can upload screenshots
or text-based chat logs, which are grouped into secure “FileBatches” linked to each user and job.
All files are fully authenticated and never visible to anyone but the uploader. Once a report is
generated, the files are immediately deleted to protect privacy and reduce storage use.
The service’s UI was designed with low-saturation, earthy tones and smooth, hand-drawn
visuals, creating a sense of warmth, trust, and clarity, even in dark mode. Reports can be
revisited and updated, and users can tap “Explore Deeper” to chat with our AI, Insight—asking
follow-up questions or clarifying context. Sometimes, Insight even asks back, creating a more
natural, reflective loop.
Through iterative testing, we shaped a workflow that’s not only technically secure, but also
emotionally safe and intuitive, especially for users dealing with vulnerable, sensitive
conversations.
