A private desktop app that runs OpenAI's Whisper locally to turn your own video and audio files into editable subtitles and language-study material in 99+ languages.
Last verified · 2026-08-28 · by Moe Ameen
SubSmith is a desktop app that turns local video and audio files into language-learning material. You drop in a file, it transcribes the speech on your own machine with OpenAI's Whisper model, and you get a timestamped transcript you can read, edit, and study from. It's built for language learners who want to immerse in real content they already have — a downloaded show, a podcast, a lecture — rather than being limited to what a streaming service happens to caption.
Because transcription runs locally, there are no uploads, no per-minute cloud fees, and no upload-size limits — the audio never leaves your device. It handles common formats including MP4, MKV, and AVI video and MP3 and WAV audio, and supports 99+ languages. Around the transcript it adds a hover dictionary for instant word definitions, one-click export of a word, its definition, and the matching audio clip to Anki flashcards, and export of the finished subtitles as an SRT or plain-text file.
SubSmith positions itself as a desktop alternative to browser tools like Language Reactor: it works on any local file instead of depending on a streaming platform or an extension. It's made by a small Birmingham, UK team and is new in 2026. Confirm current formats, platform support, and pricing on subsmith.app before relying on a specific detail.
For a creator, its usefulness is narrower and more precise than a general captioning tool: it produces an accurate, editable subtitle track and transcript for a video that lives on your drive. That's an input — a clean SRT and a word-for-word transcript — not a finished, distributed post.
SubSmith is the tool a language teacher, tutor, or polyglot creator reaches for when they want their lesson video subtitled precisely and privately: transcribe locally with Whisper, fix the timings, export a clean SRT. That SRT and the video are the raw material. [Kompozy](/) is what turns one subtitled lesson into a week of published content instead of a single upload.
Feed the finished video into Kompozy and it cuts [Clipped Shorts](/glossary/clipped-short) from the long lesson — the three or four teachable moments — and burns in captions sized for 9:16, so each vocabulary point becomes its own vertical short. The same source fans into brand-exact [Carousel Posts](/glossary/hyperframes) (a slide per phrase), Quote Graphics of the key expression, a Text Post, a Blog Article that expands the lesson for search, and an Email Newsletter for your list — every output held to one [Persona Brief](/glossary/persona-brief) so a channel of language content reads as one consistent teacher. Then [Autopilot](/glossary/autopilot) schedules and publishes the set across the eight social platforms plus blog and email behind a per-post review gate. SubSmith gets the words right in the target language; Kompozy turns that one lesson into the clips, cards, and posts that actually reach students where they scroll.
No. SubSmith runs OpenAI's Whisper model locally on your device to generate subtitles, so your files never leave your machine — there are no uploads, no upload-size limits, and no per-minute cloud transcription fees.
It works with common local formats including MP4, MKV, and AVI video and MP3 and WAV audio, exports subtitles as SRT or plain text, and supports 99+ languages. Confirm the current list on subsmith.app.
It positions itself as a desktop alternative to Language Reactor. Instead of a browser extension tied to streaming platforms, SubSmith transcribes any local file, adds a hover dictionary, and exports words plus their audio clips to Anki.
No. SubSmith produces subtitles, a transcript, and study cards — it stops at an editable SRT. To clip, caption, repurpose, and publish that video across platforms, pair it with a content engine like Kompozy.