Skip to content

Audio Insights

Transcribe audio or video with speaker diarization, then generate AI-powered summaries. Available at chat-lab.ai/audio.

How it works

  1. You provide audio through one of three input methods:
    • Upload File: MP3, WAV, M4A, FLAC, OGG, WMA, MP4, AVI, MOV, MKV, WebM, or FLV, up to 500 MB (files over 200 MB show a "large file" warning).
    • URL: a direct, publicly accessible link to an audio/video file.
    • YouTube: any YouTube URL; the audio track is downloaded server-side before transcription.
  2. Transcription: the file is sent to AssemblyAI's Speech-to-Text API with speaker diarization enabled, so the transcript is labeled by speaker (Speaker A, Speaker B, …). Most files transcribe in under a minute.
  3. Review & edit: the transcript appears in an editable text box. You can correct it, download it as Markdown, or continue to summaries.
  4. Summaries: the transcript (with your optional context notes) is sent to a large language model, which generates each summary format you selected.

Context improves summaries

Adding optional context, such as a description of the recording or details about the speakers, can improve the quality of the generated summaries.

Model providers

Task Default provider & model Optional
Transcription AssemblyAI universal speech model, speaker labels on Also supply your own AssemblyAI key
Summaries Groq serving openai/gpt-oss-120b (default) Supply your own OpenAI API key for access to GPT-5.6 family: gpt-5.6-luna, gpt-5.6-terra, gpt-5.6-sol

The free Groq tier limits how much text a single request can process, so a very long transcript may need to be edited down, or summarized with your own OpenAI key (OpenAI's models handle much longer transcripts).

User options

  • API keys (optional): supply your own AssemblyAI key (new accounts include free credits) and/or OpenAI key. Keys live only in your current browser session. Entering an OpenAI key switches summarization from Groq to OpenAI and unlocks model selection.
  • Transcript deletion toggle: when you use your own AssemblyAI key, a toggle appears under the key field. Leave it on (default) to have the app delete each transcript from AssemblyAI immediately after retrieval, or turn it off to keep transcripts in your own AssemblyAI account.
  • Contextual information: free-text notes (speaker names, topic, background) that are prepended to the summarization prompt for better results.
  • Summary formats: choose any combination of Executive Summary, Action Items / Next Steps, and Bullet Points.
  • Transcript editing: edit and save the transcript before summarizing; regenerate summaries after edits.
  • Downloads: transcript, each summary, or a combined report, all as Markdown.

Data handling

Uploaded/downloaded audio is written to a temporary server directory only for the duration of processing, deleted immediately after transcription, and swept by an age-based cleanup as a backstop. Transcripts and summaries exist only in your browser session. Audio is processed by AssemblyAI and summaries by Groq or OpenAI under their respective policies (see Privacy & Data Retention for more).

As an extra safeguard, the app deletes each transcript from AssemblyAI's servers via their API the moment it has been retrieved, well ahead of AssemblyAI's own 72-hour automatic expiry. (Uploaded audio is auto-deleted by AssemblyAI within 24–48 hours.)1 This always applies when using the app's built-in key; if you supply your own AssemblyAI key, the deletion toggle under the key field puts this under your control.

  1. Retention figures are reported as of July 2026 and were collected by Claude Code from AssemblyAI's published documentation. See Third-Party Providers for sources. ↩