Video to Notes — AI Note Generator for Any Video

Three different outputs from the same source. A transcript is the verbatim words spoken — every filler, every restart, timestamped. A summary condenses the transcript into a short overview (a paragraph or bullet list) — you can't reconstruct the meeting from it. Video notes are the structured middle ground: organized under headings, with key claims, decisions, action items, and speaker attribution — long enough to be a working document, short enough to skim in 90 seconds. Notes are what you'd hand to a colleague who missed the meeting or use as a study aid a week later. VexaScribe produces all three from a single upload — pick the format at export.

Pick by intent, not source. Study notes: for exam prep — key concepts, definitions, examples, and review questions. Meeting notes: for team calls — decisions, action items with owners, blockers, next steps. Interview notes: for user research or journalism — key quotes with attribution, themes, follow-up questions. Lecture notes: for academic content — Cornell-style with cues, main content, and a summary section. Research notes: for reading a source video you'll cite — claims, evidence, methodology gaps, and page/timestamp references. If unsure, pick meeting notes as the general-purpose default — it's the format most useful across contexts.

Different jobs. /video-summarizer produces short summaries (bullet, paragraph, timestamped chapters) — best for content you'll reference once. /video-to-notes produces long structured working documents you'll return to — best for study, research, or team distribution. /ai-note-taker is the category guide covering live-meeting bots (Otter, Granola, Fathom) — best if you want a tool that joins your meetings automatically. Rule of thumb: uploading a recording after the fact and want organized notes → this page. Live meetings with calendar auto-join → check /ai-note-taker for the right category tool.

Two accuracy layers. The transcript layer is 90-95% accurate on clean audio (single speaker, good mic, quiet room), 85-92% on Zoom/webinar recordings, 75-85% on noisy or heavily-accented content. The notes layer sits on top of the transcript — the AI reorganizes and condenses but can occasionally miss subtle emphasis, get action-item attribution wrong, or overstate the confidence of a claim. Practical guidance: treat notes as a first draft. The full transcript is included at no extra cost — verify high-stakes claims (dates, numbers, commitments) against the transcript before acting on them. For medical, legal, or journalistic use, review the transcript alongside the notes.

Yes. Paste a YouTube URL (public or unlisted) directly into VexaScribe — we fetch the video server-side, extract audio, transcribe with Whisper Large-v3, and generate notes. No download to your local machine needed. Also works with Google Drive shares, TikTok, Instagram video posts, and any direct HTTPS URL that serves an audio or video file (podcast MP3, S3 link, Dropbox share). Private, age-restricted, or members-only videos aren't accessible via URL — download those yourself and upload the file. Advantage over browser-extension summarizers: Whisper Large-v3 accuracy beats YouTube's built-in auto-captions by 5-10 percentage points on accented English and non-English content, so the notes reflect what was actually said.

Up to 5 GB and 10 hours per file. Formats: MP4, MOV, MKV, WebM, AVI, FLV, WMV for video; MP3, WAV, M4A, FLAC, OGG, AAC, AIFF, WMA, AMR, OPUS for audio. That covers 3-hour lectures, all-day workshop recordings, feature-length documentaries, 5-hour podcast episodes. For content over 10 hours, split with a free tool like LosslessCut, generate notes per segment, then merge. Free tier includes 30 minutes at signup; paid plans from $2/month for 200 minutes. A typical 60-minute video produces 800-1500 words of structured notes in 5-15 minutes total processing.

Yes — 99 languages via Whisper Large-v3. The source language is auto-detected. Notes are generated in the source language by default (Spanish lecture → Spanish notes). On paid plans you can request English notes from any source language (Japanese interview → English notes) — this adds a translation pass. Tier 1 languages (Spanish, French, German, Portuguese, Italian, Japanese, Mandarin) produce near-source-quality notes; low-resource languages produce usable but rougher output. Multilingual videos with language switching are transcribed with language boundaries preserved, then notes are generated in whichever language you request.

Markdown (best for Notion, Obsidian, Roam, Bear), DOCX (best for Google Docs, Microsoft Word, sharing with collaborators), TXT (universal fallback), and PDF (best for printing or archiving). Timestamps are preserved as clickable links back to the specific moment in the source video where each note came from — useful for study or verification. Speaker labels (when the recording has multiple speakers) are preserved through export. The full transcript is included as a separate export at no extra cost, so you can archive both the raw source and the structured notes.

Files transit over TLS 1.2+ and are stored encrypted at rest in AWS eu-west-2 (London). We do not train AI models on your videos or notes. Files can be deleted from your dashboard at any time; account deletion is self-serve. For sensitive video (internal corporate strategy, medical, legal, HR), delete the file as soon as you've exported the notes. VexaScribe is not HIPAA-eligible — do not upload PHI without appropriate contracts in place. For fully-local processing where nothing leaves your device, see /whisper-cpp or /faster-whisper.