Whishper

Whishper is an open-source, fully self-hosted transcription and subtitle suite. Upload audio or video (or paste a URL), and Whishper turns it into accurate captions with word-level timestamps using OpenAI Whisper models (via faster-whisper), all offline on a server you control. Edit the transcript in a built-in subtitle editor, then translate your subtitles with the bundled LibreTranslate engine.

Key features

• Offline speech-to-text with Whisper models (faster-whisper), word-level timestamps
• Transcribe uploaded files or media from a URL (yt-dlp built in)
• In-browser subtitle editor: fix text, re-time, and export SRT/VTT/TXT
• Subtitle translation with the bundled LibreTranslate (no third-party API)
• A small Whisper model is PRE-BAKED so first-boot transcription just works
• Co-located MongoDB bundled inside the appliance, no external database to run

Use cases

• Podcasters and video creators generating captions and transcripts in-house
• Media teams producing SRT/VTT subtitles and translations privately
• Anyone who needs offline, self-hosted transcription for sensitive audio

Why run it on the NetActuate edge Transcription of large media files means big uploads: running Whishper in a PoP near your team shortens that path, keeps your source audio and transcripts on infrastructure you own, and lets you scale CPU for faster or larger models.

Deployed from the NetActuate marketplace, Whishper boots ready to use: the MongoDB credentials are generated and persisted, the pre-baked tiny model is the default, and the web UI answers on HTTP port 8080. Larger models (base, small, medium, large) download on first use. Whishper has no built-in user login; put TLS termination and access control in front before exposing it publicly.