Karel Veselý

Lucidcast - Smart Android podcast player with on-device AI

by
Privacy-first podcast player for Android. On-device Whisper transcription, AI episode summaries, voice commands, Smart Pause. Free, no ads, no account. Made in EU

Add a comment

Replies

Best
Karel Veselý
Maker
📌
I commute by bike with Bluetooth headphones. Most podcast apps I tried would glitch — they'd skip when the BT connection dropped for half a second, or stop and never resume cleanly when I went through a tunnel. I'd lose the thread of an episode every other ride. That was the original itch. I wanted a player that just kept playing through the small hiccups: a buffer that survived weak signal patches, an auto-download that quietly grabbed the rest of the episode once it knew I was committed to listening, a Smart Pause that knew the difference between "ambulance going past" (pause) and "song in the background" (don't pause). The bike commute is also where the voice commands clicked into place. Reaching to my phone to skip a section is unsafe and annoying. So I built a wake-word system that hears "Podcast play / stop / skip / save" and responds without me touching anything. The "save thought" command I use the most — when I hear something interesting, I just say it and the app keeps a 30-second clip + transcript + AI summary I can come back to later. The next layer was about actually understanding what I was listening to. I listen at 1.7-2x — most podcast hosts speak slowly enough that 1x feels wasteful — but speeding up makes mumblers and quiet guests harder to follow. So Lucidcast got Voice Boost (auto-levels loud and quiet speakers), Speech EQ (sharper consonants for noisy environments like wind on the bike), and Skip Silence (clips out the pauses that don't carry information). Together they shave 30-40 % off most episodes without losing the thread. Features expanded organically from there. I drive on weekends and wanted live radio in Android Auto with auto-reconnect when a stream drops. I listen to long-form interview podcasts and wanted local Whisper transcription + AI summaries so I could decide what was worth re-listening to. Each feature came from a real "today this annoyed me" moment, not from a feature roadmap. The build process evolved too. The first six months it was a side hobby just for me. Then a disk crash wiped six weeks of uncommitted work (lesson learned painfully) and I rebuilt with proper discipline. Released the first public build a year in, kept iterating with real user feedback, and got to v1.1.0 this week. Privacy-first by default — no accounts, no trackers, no ads, Made in Czech Republic, one-person team, no investors.