Babel Fish AI is an open-source browser extension that I developed without writing a single line of code myself. Thanks to Roo Code, an autonomous AI agent, I created an efficient voice transcription and translation solution, while exploring technologies like Heygen, FFmpeg, and DALL-E to enhance the project. It’s not just a technical tool: it’s a learning journey, a technology watch effort, and an attempt to provide a quality user experience for an international audience.
Features
- Voice transcription: High-quality audio capture, accurate transcription via Voxtral (Mistral AI) or Whisper (OpenAI)
- Real-time translation: Optional translation of transcriptions into 15 languages
- Context menu: Rephrase, translate, or correct the spelling of any selected text
- Multi-provider: Mistral AI, OpenAI, or a custom endpoint (LiteLLM)
- Multi-browser: Chrome and Firefox supported
Developing with AI: An Innovative Approach
Babel Fish AI is entirely based on Roo Code, which generated every line of code – from the Chrome manifest to scripts and the interface. Without skills in CSS, HTML or JS, I guided the AI with precise prompts. When it hit obstacles, I dove into the documentation to help it move forward, proving that with method you can overcome technical barriers. I tested several AI models via Roo Code: initially exploring their capabilities, then opting for Gemini (notably Gemini 2.0 Pro Experimental) for its free cost and good value. I also used Claude Sonnet, which proved better optimized for certain tasks in Roo Code. Choosing the right model at the right time was key for this project.
Evolution: From Roo Code to Coding Agents
Since late 2025, I have maintained Babel Fish AI with coding agents like Claude Code and Gemini CLI. This continuity in approach allowed me to add many features: Firefox support, integration of Mistral AI as a transcription provider (via Voxtral), the context menu for rephrasing and correcting text, and many other improvements.
Multi-Browser Support
The extension is available on Chrome and Firefox:
Multilingual Videos: An Ambitious Multimedia Effort
To showcase Babel Fish AI, I created demo videos in several languages with Heygen. The process was intense: I captured footage with OBS Studio, extracted audio with FFmpeg, then assembled everything in Heygen. To add a personal touch, I generated a unique avatar with DALL-E, which brings those videos to life. This work, although time-consuming, helped make the extension accessible to a global audience, and I loved taking on this multimedia challenge for the first time.
Translation and International Accessibility
Thanks to Roo Code, I translated the extension interface into many languages using well-crafted prompts. What would have taken hours manually was completed quickly with AI, strengthening Babel Fish AI’s accessibility. It was also an opportunity to explore software localization, a field I had never touched before.
Quality and Professionalism
Rigor has been at the heart of this project:
- Tests and Tools: I used code scanners like SonarCloud, CodeFactor, Codacy, and DeepSource to check the quality of AI-generated code, in addition to thorough manual testing.
- Privacy Policy: I wrote a detailed privacy policy, a sign of transparency and seriousness for users.
- Project Management: Everything is versioned on Git with clear pull requests, ensuring clean and organized tracking.
These efforts show that even without coding directly, it’s possible to deliver a reliable and professional product.
Open Source and Sharing
Babel Fish AI is distributed under the GNU Affero General Public License v3.0. This choice reflects my belief that technical knowledge should be shared. The source code is open to contributions, and the README is translated into 14 languages to welcome an international community.
A showcase site presents the project and its features.
Technology Monitoring and Learning
This project opened doors to tools and skills I had never used before:
- AI coding agents (Roo Code, then Claude Code and Gemini CLI)
- Video creation with Heygen, OBS Studio and FFmpeg
- Avatar generation with DALL-E
- Drafting legal documents like the privacy policy
- Using code analysis tools to ensure quality
It was a real technology watch session, and I learned a lot by diving into these new areas.
Usability and User Experience
Even without design expertise, I developed a clear vision of usability (UX/UI). I guided Roo Code to create an intuitive interface, with choices like a well-integrated context menu and clear transcription results. This focus on user experience demonstrates my attention to detail and my desire to meet real user needs.
Try Babel Fish AI!
- Official site: babelfishai.jls42.org
- Source code: GitHub
- Chrome: Chrome Web Store
- Firefox: Firefox Add-ons
What This Project Says About Me
Babel Fish AI is more than an extension: it’s a showcase of my ability to innovate, learn, and deliver a finished product. I managed to leverage AI as a collaboration tool, handle a complex project, and explore new technological horizons, while maintaining high standards. This journey reflects my curiosity, rigor, and desire to create useful tools and share them with the community.
This document was translated from the fr version into the en language using the gpt-5-mini model. For more information about the translation process, see https://gitlab.com/jls42/ai-powered-markdown-translator