Week 5 #

Project name: VoiceDiary

Code repository: link

VoiceDiary is an AI-powered voice journaling tool that analyzes tone, emotion, and key themes in spoken entries. It generates personalized emotional insights and well-being suggestions based on recorded reflections.

Team #

Feedback #

Sessions #

1. "Convenient and beautiful interface, intuitively clear where to click. The site is gorgeous. The output wasn’t very clear to me, but I used it just for fun—it was interesting to see something come out. It’s fascinating to observe how the neural network reacts to my input. I haven’t tried the voice AI yet, but it seems intriguing."

2. "I visited the site—it’s beautiful, and at first glance, it seemed like something serious. I tried speaking into it, and while the topics were somewhat relevant, the psychological insights didn’t quite fit me. The results were in English even though I spoke in Russian. I’d use it more if the app were higher quality."

3. "The concept is cool, and the design is modern, but the responses sometimes don't match. I would like if it could all be edited, because I would like to use it daily, sometimes the emotion is not accurate. Also, adding multilingual support would make it more accessible."

Analize #

Strengths #

✅ UI/UX praised – Users highlight the intuitive, visually appealing interface.
✅ Engagement potential – Some interact “for fun,” showing interest in observing AI behavior.
✅ Voice input curiosity – Ineresting feature.

Pain Points #

⚠ Output clarity issues – Confusion over results (e.g., psychological insights feeling mismatched).
⚠ Language mismatch – Russian input yielding English outputs reduces usability.
⚠ Perceived lack of depth – Feedback suggests responses can feel low-quality.

Actionable Improvements #

High Priority #

🔧 Customizable output fields – Add emotion editing possibility (to correct when the model is wrong)

🔧 Input-language consistency – Warn users that product functionality is in English.

Mid/Long-Term #

🌍 Multilingual support – Currently high effort; requires deeper pipeline changes.
🤖 Emotion recognition improvement – Refinine context retention, improve emotion recognition.

Additional Notes #

• Some users treat the tool as entertainment—could lean into this with a “playful mode” variant.

Iteration & Refinement #

The product is fully functional and stable in production. While we’re actively addressing minor issues, the core system performs reliably. Next week, we’ll implement several researched features to enhance functionality.

📌 Key Progress Highlights: #

Performance & Stability: Rigorous testing has ensured a robust foundation. Remaining issues are edge cases (e.g., rare latency spikes) being prioritized for fixes.

ML Model Refinement:

    Emotional Insights: Discovered a more accurate approach for emotion analysis (higher precision/recall).

    Orchestration: Finalizing multi-model pipelines for end-to-end emotional recognition.

Documentation:

    Structured Guides: READMEs in /frontend and /ml detail setup, architecture, and workflows.

    Experimental Work: /experimental includes R&D notes for upcoming features.

Next Steps: #

Deploy refined emotional recognition models.

Optimize API handoffs between backend/ML services.

Individual commitments #

Plan for Next Week #

ML #

• Finish emotion recognition model orchestration

Frontend #

• Do a manual emotion change function
• Collaborate with backend

Backend #

• Enhance frontend integration
• New user-facing features to edit generated fields
• Expand test coverage for the Go service
• Optimize CI/CD pipeline

Confirmation of the code’s operability #

We confirm that the code in the main branch:

• [✓] In working condition.