Overview
Lucent Flow is a web meditation experience where users describe what they need — typed or spoken — and receive a session generated for that moment. The product interprets emotional context and preferences, then builds guidance through progressive audio segments, natural pacing, and an immersive player designed to feel calm and distraction-free.
Stack
Next.js · React · OpenAI (intent & narration structure) · ElevenLabs (voice) · Supabase (sessions, segments, favorites)
Core systems
Intent interpretation · segmented generation · ambience layering · replay vs regenerate · AI validation for coherence and pacing
Design focus
Conversational composer · subtle customization · immersive playback · light/dark calm UI aligned with portfolio design tokens
Context — why intent first
Most meditation apps organize prerecorded libraries by category. Lucent Flow explores whether a session can begin with what the user actually feels — wind-down after stress, focus before work, a short breathing reset — and adapt technique, tone, and duration from that moment.
Intent-based generation
Natural language (or voice) is structured into goal, emotional tone, duration, voice, technique, and ambience. The interface stays conversation-first; advanced controls remain available but secondary.
Progressive session construction
Sessions are built in intro, core, and closing segments so playback can start before the full session exists. The player tracks whole-session progress while duration estimates refine as new audio is generated.
Pacing, silence, and believability
A major focus was variable pauses, breathing alignment, and technique-specific rhythm — slower sleep pacing versus steadier focus continuity. Small inconsistencies break immersion, so pacing became a central design and engineering problem.
AI validation layer
Generated guidance is checked for logical continuity, breathing coherence, conflicting instructions, and repetitive phrasing — with regeneration when needed. This is framed as product quality, not novelty for its own sake.
UX evolution
The interaction model moved from early technical controls toward replay-locked saved sessions, edit-to-create-variation, favorites and recent memory, and a cinematic player that shares typography, icons, and theme behavior with the portfolio ecosystem.
Immersive playback
A distraction-free layout with ±10 second skip controls, favorites, session-level timing, and calm light/dark modes — intentionally not a generic podcast UI.
Outcome
Lucent Flow combines conversational AI, adaptive audio systems, and emotionally intelligent interaction into one coherent product — both a UX exploration and a technically demanding orchestration challenge.
Deployment note
Portfolio: daniel-ml.com. Live app (default): lucentflow.daniel-ml.com. Launch links resolve via src/config/lucentFlowAppUrl.ts — set PUBLIC_LUCENT_FLOW_APP_URL for staging or custom hosts; local dev defaults to http://localhost:3000.
