Paste a YouTube link get the TLDR + chapters back.

The transcript is the full word-by-word text — useful when you want to read or quote precisely. The summary is the distilled version — TLDR, chapters, key quote — useful when you want to decide if the video is worth watching, or to write a blog post that references it. Both come from the same run.

Yes. Default is a 5-bullet TLDR (~80 words). 'Short' gives you a single tweet-length sentence. 'Long' gives you a ~200-word paragraph. 'Exhaustive' gives you a section-by-section walkthrough — every chapter expanded into 3-4 sentences. Same transcript, different summarization pass.

The transcript itself is Whisper-large-v3 accuracy (~95% on clean audio). The summary is LLM-generated from that transcript. The 'key quote' is an exact substring of the transcript with a timestamp — those are reliable. The bullet points are paraphrased — verify any specific claim against the transcript before quoting in print.

Honest answer: it can. We've reduced hallucinations by grounding everything in the transcript and citing timestamps, but LLM summaries are not infallible. For any high-stakes claim (data, a person's exact words, a contested fact), verify against the timestamped transcript before publishing.

The transcript auto-detects 100+ languages. The summary works best in English, Spanish, French, German, Portuguese, Italian, and Hindi. For other languages the transcript is solid but the summary may be less precise — verify against the transcript or run a translation step.

Two to four minutes for a one-hour video on a clean recording. The transcription step is the bottleneck; the summary itself is generated in under 10 seconds once the transcript is ready. Live progress shows in the queue.