CONTEXT AND BRIEF
Dream journaling has a retention problem. Everyone who tries it quits within weeks. The friction is not motivation. It is the moment of capture.
Onyric is a personal project exploring a design problem I kept running into in conversation: dream journaling is a practice that almost everyone who tries it abandons within a few weeks. The interest is genuine. The habit does not survive contact with the moment of capture. Preliminary conversations with eight colleagues who had all attempted dream journaling at some point confirmed the pattern. The failure was consistent enough to be worth investigating as a design problem rather than a motivation problem.
Every dream journal app in the market followed the same paradigm: open the app, type what you remember. The problem is that typing coherent narratives immediately after waking requires cognitive effort at exactly the moment when dream memory is most fragile and the user is least capable of it. Six of twelve people interviewed had tried using voice memos on their phone as a workaround but found replaying audio files to review entries too tedious to sustain. The existing apps treated dreams like diary entries and assumed a fully alert user. The actual user is half-asleep, disoriented, and losing the memory with every second they spend navigating a UI.
Client
Personal project
Industry
Consumer – B2C · Mobile App
Year
2025
Role
Solo UX/UI Designer
Duration
4 months
PROBLEM DEFINITION
74% of people who try dream journaling quit within a month. Not because they lose interest. Because the moment of capture requires the one thing a person who has just woken up does not have: cognitive effort.
A survey of 47 people across dream journaling communities confirmed the retention problem: 68% had tried journaling and 74% of those had abandoned it within a month. When asked about preferred input methods, 53% chose voice as their first preference, yet no app in the category made voice the primary interaction. Twelve semi-structured interviews revealed the texture behind those numbers.
Nine of twelve people mentioned difficulty typing immediately after waking as the main friction point. Six had tried phone voice memos as a workaround and found reviewing audio files later too cumbersome to sustain.
Ten wanted to see patterns in their dreams over time but found manually reviewing entries too effortful to do consistently. One quote shaped the direction more than any data point: "If I could just talk into my phone and have it written down for me, I'd actually do it." The technology existed. No app had made it the default.
PROCESS AND EXPLORATION
The design context for Onyric was unlike anything I had worked with before. Every other project starts from a user who is alert, capable, and able to read instructions. Onyric's critical moment happens when the user is semi-conscious, disoriented, and losing the content they want to capture with every second they spend navigating. That constraint shaped every decision before a single screen was drawn.
Three home screen variations were tested via Maze with six participants. Version A placed text input first with a small voice button in the corner. Version B put the voice button at the centre with text as a secondary option. Version C made the entire screen a voice activation surface. Version B won on task completion speed, averaging 3.2 seconds to first tap against 5.1 for Version A. Version C confused participants who expected tapping to navigate rather than record.
One decision took longer to resolve than expected. Should transcription happen in real time while the user speaks, or after recording ends? Both were prototyped. Real-time transcription created anxiety: users kept glancing at the text to check whether the recognition was working, which pulled their attention away from the memory they were trying to preserve.
Post-recording transcription let them speak without interruption. A clear loading state with the label "Transcribing your dream..." replaced the ambiguous "Processing..." that three early testers found confusing. The information architecture was deliberately minimal. Three screens: capture, timeline, insights. Nothing else. Every feature that did not serve one of those three jobs was removed during the process, including a secondary tab on the Insights screen that tested poorly and added no value for the core use case.
Chose post-recording transcription over real-time. Users speaking freely into a phone produce better content than users monitoring a text field for errors while simultaneously trying to remember a dream.
Designing for a semi-conscious user immediately after waking. Zero tolerance for friction, navigation complexity, or instructions. Every interaction had to work without the user being fully alert.
A voice-first app designed for one very specific user: a groggy person at 6am with a dream evaporating in real time. You speak. Onyric does the rest.
Prototype test participants completed the recording task without prompting
7 / 8
Average time to first tap in usability testing (scenario: just woken from a vivid dream)
2.1 sec
SUS score post-testing, above the 68-point industry average for usability
78.1
