Vela is a memory companion for iPhone. Add the people you love, listen to conversations together, and recall what you talked about — all without a single byte leaving the device.
A daily companion for navigating conversations when names, faces, and details start to fade.
Add the people who matter — moms, dads, doctors, friends. Each profile keeps a photo, relationship, voice sample, and a stream of past conversations.
Hit Listen and Vela turns the conversation into real-time text on your iPhone screen. Large font, high contrast, designed for hearing loss too.
Vela separates speakers in the conversation. Tap "Speaker 1" and assign it to Mom — every line she speaks gets her name from then on, this conversation and forward.
Conversations save automatically with title, date, participants, and full transcript. Search later by person, keyword, or date. Re-read what was said.
Record 10 seconds of someone's voice and Vela uses it as the signature for future tagging. Optional — manual tap-to-tag works too.
Apple Neural Engine + GPU run every model inside your iPhone. No audio, no transcripts, no photos ever leave the device. No accounts, no cloud, no analytics.
Voices, faces, and the things your loved ones said over a cup of coffee live only inside your iPhone. After the initial model download, Vela never connects to the network. Uninstall the app and every conversation, every face, every voice sample disappears with it. We never see any of it.
The same wave that drove ClearCaps — adult children doing their best for the parents who raised them.
"Mom forgets which doctor said what. Dad asks who that person was three times during dinner. Vela won't fix what's underneath — but it can keep the names, the faces, and the warm small conversations close. That's worth a lot."
AstroLexis builds mobile tools to give people solutions. Vela is one of them.
No. Vela is an accessibility tool. It is not a medical device, is not approved by FDA or any health authority, and does not diagnose, treat, or replace medical care for dementia, Alzheimer's, or any cognitive condition. It is a complementary aid for daily life.
An on-device diarization model (Pyannote via SpeakerKit, running on the iPhone's GPU) splits the audio into separate speaker clusters during the conversation. You tap "Speaker 1" → pick a person from your circle, and every line attributed to that cluster is relabeled. Cross-session voice matching is on the roadmap.
Three AI models run inside your iPhone: Whisper (speech-to-text, ~630 MB), Pyannote (speaker diarization, ~50 MB), and Gemma 4 (optional summaries, ~1.5 GB). All three live in your app's private storage after a one-time download. After that, audio is captured, transcribed, and stored — without any network connection. You can verify in Settings → Models.
iPhone with iOS 26 or later. For all features (especially summaries) we recommend iPhone 15 Pro or newer with 8 GB of unified memory. On older 6 GB models, the live captions and saved memories work fine — AI summaries may take longer.
Pricing not finalized. Likely in the $4.99–$9.99 range as a one-time purchase, no subscriptions, no in-app purchases. Consistent with the rest of AstroLexis's iOS apps.
Vela is built by AstroLexis — a small company building mobile tools to give people solutions. The same team built ClearCaps (live captions for hearing loss). Contact: contact@astrolexis.space.
