@Firebase @FlutterDev Gemini TTS Use Case: "Finnish It" Finnish It is a coaching app designed to help adult immigrants in Finland prepare for their national language certificate (YKI), which is heavily used for job and citizenship applications. I recently used the Firebase AI Logic SDK to implement a feature that simulates the intermediate-level paired conversation task ("Keskustelu"). In the actual exam, candidates interact with a pre-recorded partner in a language lab: 🔵 They are given a scenario (e.g., returning broken goods to a store) 🔵 A pre-recorded audio of the partner plays 🔵 A beep sounds, and the candidate gets an imperative prompt on-screen (e.g., "describe the problem and ask for help") with a strict 15-to-30-second speaking window. They never see the partner's script, because the task is listening comprehension as much as speaking. To simulate this authentically, the app uses @GeminiApp Flash model to dynamically generate the full exercise structure. Model generates the partner's turns, the imperative instructions for the candidate, sample responses, and specific duration constraints for each turn. This is tailored to the learner's CEFR level and the exact, personalized scenarios they want to study, ensuring they are ready for any situation before the exam. Finally, @FlutterDev firebase_ai package uses the Gemini TTS model to synthesize the partner's dialogue.