Speji — Real-time translation overlay for Android

Capabilities

How Speji works

Like X-ray glasses for foreign languages. Three steps. Zero friction.

OS-level audio capture

Speji uses Android's official audio-capture API — no app-specific integrations needed. It picks up audio from compatible apps (short-video, music, games, and more). Video-call apps and some streaming services block sharing and won't be captured.

Real-time translation

A low-latency pipeline transcribes and translates speech as it streams in. Subtitles appear word by word, with sub-second delay on a good connection.

Transparent Overlay

Subtitles float transparently over your content. Like the glasswing butterfly it's named after — barely there, beautifully clear.

A growing language set

Speji ships with a curated list of languages — more on Standard, even more on Pro — and we add new ones on a rolling basis. Some are available as translation targets only, depending on speech-recognition quality.

Lightning fast

Audio is streamed to fast speech and translation engines through our low-latency backend. Sub-second from speech to subtitle on a good connection.

Private & Secure

Audio is processed in real-time and never stored. Your conversations and content stay private. Always.

Our Origin

Where it started

We live in an era of AI abundance. More content than ever before — anime, documentaries, tutorials, memes — created in every language, accessible to anyone with a screen. But understanding it? Still broken.

You're watching an anime and the official translation takes months. You find an educational video on YouTube in Japanese, or a meme on TikTok in Portuguese that's genuinely funny — but you can't fully enjoy it, appreciate it, or share it, because you don't understand the words.

There's no lack of content. There's a barrier. And that barrier doesn't need to exist anymore.

Speji is built on one idea: language should be transparent. Every piece of audio content on your device should be accessible — instantly, privately, without switching apps or waiting for a human translator.

Why "Speji"?

In the rainforests of Central and South America, there lives a creature that defies the rules of nature: the Greta oto, or Glasswing Butterfly.

Unlike other butterflies that use bright colors for camouflage, the Glasswing has evolved a different superpower: Transparency. Its wings are so clear, so free of pigment, that they are almost invisible. In Spanish, they are affectionately called espejitos — or "Little Mirrors."

That's Speji. A transparent layer that sits over your screen — barely there, beautifully clear — and instantly reflects meaning back to you from any audio, in any language.

The Team

Built by people who needed it — for themselves

No product brief, no market research. Just engineers who hit the same wall every day and decided to break it.

Engineers first

We build software for a living. Speji was built the way engineers build things for themselves — precise, minimal, and without unnecessary noise.

Zero interest in your data

We process audio in real-time and discard it immediately. We built Speji because we hate data collection as much as you do. No logs, no training sets, no profiles.

Building something new

There's no shortcut to real-time, on-device audio translation across any app. We built the infrastructure from scratch because what we wanted didn't exist yet.

Questions

Frequently asked

Quick answers about how Speji works, what's in the beta, and where your data goes.

Which apps does Speji work with?

Speji uses Android's official audio-capture API, so it works with apps that allow their audio output to be shared with other apps — most short-video, music and game apps fall in this category. Video-call apps don't expose their audio to other apps and are not captured. A few streaming apps also block capture depending on how they route audio. If an app's audio can't be captured, Speji simply won't pick it up — nothing breaks.

How is my audio handled? Is anything stored?

Audio is processed in real time and discarded immediately. We don't store audio, build user profiles, or use your conversations to train models. See our privacy policy for the full breakdown.

Which languages does Speji support?

Speji ships with a curated language set that grows over time. The exact list depends on your plan — the Free tier covers a small core (currently English, Spanish, French), and Standard and Pro unlock additional source and target languages. Some languages are available only as translation targets, where speech-recognition isn't yet at the quality bar we want. We add new languages on a rolling basis.

How can I get into the beta?

Speji is in closed Android beta. Request a tester slot from the early-access page; we approve testers manually within 48 hours and send a one-tap install link. iOS is on the roadmap.

Is Speji free?

Yes. Speji has a free tier with daily limits. A Pro plan unlocks longer sessions, more languages, and ad-free use. Beta testers help us shape both tiers.

Does it work offline?

Real-time translation needs a network connection so we can route to fast speech and translation models. Offline mode is on the long-term roadmap.

What about copyright and platform terms?

Speji is a personal accessibility and learning tool. You're responsible for using it only with content you're entitled to access and in compliance with each platform's terms of service and applicable copyright law. Speji does not authorize or endorse any use that breaches a platform's policies or copyright, and we are not responsible for misuse.

Still have questions? Contact us

See through every language barrier