AudioTextify
Ninety-nine languages in a single browser tab. Click a button and the audio becomes text you can read, summarize, translate, or turn into new content.
Industry: Productivity, Multilingual Content | Stack: Chrome MV3, TypeScript, Speech APIs, Translation APIs, LLM Integration | Status: Live on Chrome Web Store | Chrome Web Store
what AudioTextify does#
AudioTextify is a Chrome extension that runs audio and video content through four AI operations: transcription, summarization, translation, and content generation. Open a French video lecture, click the extension, and you get an English transcript. From there you can summarize it, translate it into a third language, or generate a blog post from it. The whole chain runs inside the browser tab.
The extension supports 99 languages for transcription and translation, processes MP3, WAV, and video formats, and splits into free and premium tiers based on monthly usage volume. Premium gives you 10 hours of processing per month.
the multilingual challenge#
Building a transcription tool is straightforward. Building one that holds up across 99 languages is not.
The problem compounds fast. A speech recognition model trained mostly on American English will stumble on Indian, Scottish, or South African accents. Tonal languages like Mandarin and Vietnamese need a different approach entirely. Now add the downstream work: after transcription, the system needs to summarize or translate, and that output has to read like actual language, not a mechanical word swap. Context, idiom, and domain-specific terminology all require handling that a naive pipeline gets wrong.
The client was French, which meant this was not an abstract multilingual requirement. He was building for users who actually work across languages every day, and the standard tools either required file uploads, context switching, or just did not support the language mix he needed. The constraint he gave us, staying inside the browser without making users leave the page, shaped every decision from there.
Chrome's Manifest V3 does not make this easy. Service worker lifetimes are enforced. Background processing is restricted. Persistent storage is limited. Running four sequential AI pipelines (speech-to-text, NLP summarization, translation, content generation) inside those bounds requires careful API orchestration, state management across the service worker lifecycle, and fallback behavior when rate limits or network conditions intervene. Getting it to feel instant when the user clicks the button took more work than the AI integration itself.
what we built#
The core is a multi-stage pipeline where each operation feeds the next. A user can stop at transcription, or keep going through summarization, translation, and generation, with each stage using the output of the previous one.
The transcription layer handles 99 languages at 90%+ accuracy on clear audio and detects the source language automatically when the user does not specify it.
Summarization condenses raw transcripts into structured output: key points, main arguments, and takeaways. A 45-minute lecture produces thousands of words of transcript; the summarization layer makes that usable without the user having to read all of it.
Translation goes further than word substitution. Technical terms, idiomatic expressions, and domain-specific language all need context-aware handling. The output is meant to read naturally in the target language, not like it was run through a dictionary.
The final stage generates new content from whatever was processed. A researcher working from a foreign-language interview can get an English article draft. That is the case that sold the client on the project.
On the infrastructure side, the extension runs on Manifest V3 with service worker scripts, content scripts for audio detection and extraction, a popup and side panel UI, an API orchestration layer handling sequential and parallel calls, and Stripe for subscription management.
key capabilities#
- One-click transcription: Transcribe audio and video content directly in the browser without uploading files to a separate service
- 99-language support: Transcription and translation across 99 languages with automatic source language detection
- AI summarization: Condense long transcripts into structured summaries with key points and takeaways
- Cross-language content generation: Generate new content in any supported language from audio sources in any other supported language
- Free and premium tiers: Free plan with 30 minutes/month of processing, premium at $9.99/month with 10 hours/month
- Editable outputs: All generated transcripts, summaries, and translations are fully editable before export
results#
The extension shipped to the Chrome Web Store and is actively maintained. The French client got a tool that covers his full language mix, handles the in-browser constraint, and supports a four-stage workflow from raw audio to publishable content. The build covered everything from initial architecture through store publication.
Building a multilingual AI tool or Chrome extension? See our Browser Extension Development service or book a free Automation Audit.
