SPECTER — Offline AI Subtitle Generator for Android
Offline subtitle generation for Android
Import audio, video, or a ZIP archive. SPECTER generates one subtitle file per media file on your device.
No account. No cloud upload. No transcription server.
On-device transcription
Processing stays on your phone
Audio, video, ZIP input
MP3, MP4, MKV, WAV, and ZIP archives
One SRT per file
One subtitle file per source file
No account required
No login. No signup. No cloud.
How it works
IMPORT
Drop audio, video, or ZIP files
PROCESS
On-device AI generates subtitles
EXPORT
One .srt file per source file
See it in action
Technical specifications
SUPPORTED_FORMATS
| TYPE | FORMATS |
|---|---|
| Audio | MP3 WAV AAC OGG FLAC WMA M4A OPUS |
| Video | MP4 MKV AVI MOV WEBM FLV WMV 3GP |
| Archive | ZIP |
| Output | SRT |
AI_MODELS
| MODEL | SIZE | RAM | SPEED | TIER |
|---|---|---|---|---|
| tiny | ~75 MB | ~1 GB | Fastest | Basic |
| base | ~150 MB | ~1.5 GB | Fast | Standard |
| small | ~500 MB | ~2.5 GB | Moderate | Quality |
| medium | ~1.5 GB | ~4 GB | Slow | High |
SYSTEM_REQUIREMENTS
| PARAMETER | REQUIREMENT |
|---|---|
| Platform | Android 8.0+ (API 26) |
| Storage | 200 MB + model size |
| RAM | 2 GB minimum (4 GB+ recommended) |
| Processor | ARM64 (64-bit required) |
Privacy guarantees
No data transmitted during transcription
All processing on-device
No analytics, no telemetry, no cookies
SPECTER operates entirely within your device boundary. No network connections are made during transcription. Audio data never leaves the phone. There are no user accounts, no usage analytics, and no third-party SDKs.
Download
Technical dossier
Processing pipeline
IMPORT
Load media from device storage
EXTRACT
Isolate audio channel from container
INFER
Run Whisper model on audio segments
GENERATE
Assemble timestamped SRT entries
EXPORT
Write .srt file to output directory
Design system architecture
SPECTER uses two sibling design systems: PHOSPHOR for the Android app and SIGNAL for this site. Both share a three-layer token architecture that maps raw values through semantic roles to component-specific aliases.
--color-spectra-400: oklch(0.66 0.18 280) --text-accent: var(--color-spectra-400) --card-icon: var(--text-accent) Engineering constraints
The entire transcription pipeline runs without network access. The Whisper model ships with the APK. No server, no API key, no fallback to cloud.
Audio processing and speech recognition execute on the device GPU/CPU. Model selection adapts to available RAM, from 1GB base to 3GB turbo.
No network permission is requested. No telemetry SDK is included. No user data leaves the device boundary at any point in the pipeline.
Frequently asked questions
Does SPECTER upload my media?
No. All processing happens on your device. No audio, video, or transcript data leaves your phone.
What files does it support?
Audio (MP3, WAV, M4A, AAC, FLAC, OGG, OPUS, WMA), video (MP4, MKV, MOV, AVI, M4V, WebM, TS, FLV), and ZIP archives containing up to 50 files.
Is Google Play or APK download safer?
Google Play is the recommended path. It handles signing verification automatically. APK install is for advanced users who want to verify the release themselves.
Does it work offline after install?
Yes. SPECTER works fully offline, including airplane mode. The bundled AI model is included with the app.
Which languages are supported?
The Whisper AI model recognizes speech in 99+ languages. You can transcribe in the original language or translate to English.
Can it translate subtitles to English?
Yes. SPECTER supports translate-to-English mode for non-English source audio.
Why does the app need storage access?
To read your media files and save the generated subtitle files. No other file access is required.
Is this website the portfolio?
No. This is the SPECTER product site. It is designed to be strong enough to appear as one piece of portfolio work.