Upload a ZIP from interview_recorder.html. The page transcribes each audio answer with Gemini, asks an LLM for a structured fluency analysis in the student's native language, and produces a downloadable HTML report styled like new_student_intake.html.
1Student details & ZIP upload
2Transcribe audio
Ready to transcribe.
3Generate fluency analysis
Sends all transcripts plus the question text to NVIDIA's Llama 3.3 70B and asks for a structured assessment in the student's native language. Takes about 5 minutes — keep this tab open while the six stages run:
Hero & CEFR placement — picks the student's level (A1–C2) and writes the warm overview.
Fluency commentary — four paragraphs on overall fluency, pacing, vocabulary, and interaction.
Strengths & growth — six concrete things the student does well, six gentle next steps.
Beautiful sentences — five real sentences from the transcripts, diagrammed with POS tags and phrase brackets.
Pronunciation focus — two multi-syllabic words from the answers that are tricky for the student's L1, plus an AWS Polly native reference clip for each.
Encouragement letter — a warm, personal closing letter signed "Your English teacher".