Meeting Transcript Summarizer

Overview

Summarizes a meeting from a recording or from a pasted transcript, and returns action items as table rows instead of prose.

1. Inputs - Recording (nodetool.input.AudioInput) and Transcript (nodetool.input.StringInput, multiline). Fill in either one. The template ships with a sample recording, so clear it when you paste a transcript.
2. Routing - A Code node tests whether the pasted transcript is empty. Two If nodes pick the transcribed audio when it is empty and the pasted text otherwise. A second Code node joins the chosen text.
3. Transcription (nodetool.text.AutomaticSpeechRecognition) - Transcribes the recording with openai/whisper-large-v3 on fal.ai.
4. Summary (nodetool.agents.Summarizer) - Writes a Markdown summary of about 150 words with Overview, Key discussion points, and Open questions.
5. Action items (nodetool.text.Prompt into nodetool.generators.DataGenerator) - Returns a table with action, owner, and due_date columns.
6. Outputs - Transcript, Summary, and Action Items.

Demo

Tags

audio, llm, dataframe

Workflow Diagram

graph TD recording["AudioInput (Recording)"] pasted["StringInput (Transcript)"] asr["AutomaticSpeechRecognition"] empty["Code (is empty)"] useAudio["If (use transcribed audio)"] usePasted["If (use pasted transcript)"] join["Code (join)"] summarizer["Summarizer"] prompt["Prompt (action items)"] generator["DataGenerator"] transcriptOut["Output (Transcript)"] summaryOut["Output (Summary)"] actionsOut["Output (Action Items)"] recording --> asr pasted --> empty empty --> useAudio empty --> usePasted asr --> useAudio pasted --> usePasted useAudio --> join usePasted --> join join --> summarizer --> summaryOut join --> prompt --> generator --> actionsOut join --> transcriptOut

Edit this page on GitHub