All videos
0:00 / 0:00
research

Clicky Meets Universal-3.5: Voice-Navigate Your Code

AssemblyAI30 June 2026Watch on YouTube

Part of series

Ep. 2 · Universal 3.5: Spraakherkenning in actie

Demo's en tutorials van AssemblyAI's Universal 3.5 Pro speech-to-text model met contextuele awareness en real-time transcriptie.

View the series

What you'll learn

  • Dynamic context-prompting significantly improves speech recognition accuracy for technical terminology
  • AssemblyAI's Universal-3.5 model automatically recognizes relevant code terms, APIs, and identifiers based on what's visible on screen
  • Voice-based code navigation becomes practical when developer-specific language is recognized accurately and natural questions like 'Where is PCMS16LE?' are interpreted correctly
  • Contextual prompting is an effective technique to make speech interfaces and AI assistants more accurate for technical workflows

Frequently asked questions

How does dynamic context-prompting improve speech recognition for code?
The system automatically inserts relevant terms, APIs, and identifiers based on what the developer currently sees on screen. This gives the model much better context to recognize technical terminology accurately.
Which speech recognition model is used in this demo?
The demo uses AssemblyAI's Universal-3.5 model, integrated with Clicky for voice-based code navigation.
What types of applications is this relevant for?
This is relevant for building developer tools, voice interfaces, and AI assistants where accurate recognition of technical language and natural questions about code is essential.
What are examples of questions users can ask?
Users can ask questions like 'Where is PCMS16LE?' or 'What is audio PCM16 data doing?' and the system navigates to the correct code.

Topics

speech-recognitionai-modeldeveloper-toolsnlpmultimodal-ai

Read next

Description from the channel

Technical words are hard for speech recognition—but they don't have to be. This demo shows how Clicky integrates AssemblyAI's Universal-3.5 model with dynamic key-term prompting. As the user scrolls through a codebase, Clicky automatically updates the transcription context with relevant identifiers, APIs, and technical terminology currently on screen. The result is significantly more accurate recognition of developer-specific language, allowing users to ask natural voice questions like "Where is PCMS16LE?" or "What is audio PCM16 data doing?" and instantly navigate to the correct code. Whether you're building developer tools, voice interfaces, or AI assistants, this is an example of how contextual prompting can improve transcription accuracy for technical workflows. ▬▬▬▬▬▬▬▬▬▬▬▬ CONNECT ▬▬▬▬▬▬▬▬▬▬▬▬ 🖥️ Website: https://www.assemblyai.com 🐦 Twitter: https://twitter.com/AssemblyAI 🦾 Discord: https://discord.gg/Cd8MyVJAXd ▶️ Subscribe: https://www.youtube.com/c/AssemblyAI?sub_confirmation=1 🔥 We're hiring! Check our open roles: https://www.assemblyai.com/careers ▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬ #MachineLearning #DeepLearning