Loading sample…
Video to Markdown API
Turn a video URL into timestamped Markdown.
vid2md pulls a video, processes it on transient Cloudflare container disk, writes Markdown, then deletes the video working files.
NEW Powered by google/gemini-3.5-flash — the latest from Google.
Try it now · free 60s
Convert your first 60 seconds.
Paste a URL or upload a file. Ready in 60–90 seconds. Email unlocks the result.
Free demo
Generating Markdown
Extracting keyframes, transcribing audio, writing Markdown.
See it work
Real video in. Real Markdown out.
Three flavors of input. Same concise mode. The output on the right is what vid2md actually returned.
How people use it
Four scenarios, one output.
vid2md takes any video and returns Markdown with what was said and what was shown, aligned to timestamps. Four real workflows it replaces:
Coding agents
Visual context for agents
Who: engineers running Claude Code, Cursor, or a custom agent loop.
Input: 4-minute screen recording of a failing UI flow, detailed mode (4×).
Output: Markdown with frame observations at every state change (modal opens, network tab flashes red at 02:14, console error at 02:16), narration transcript, final summary.
Agents read Markdown, not MP4. One file replaces 40 screenshots and a written repro.
Support
Loom bug reports into tickets
Who: support engineers triaging customer Loom links.
Input: Loom URL, concise mode (1×).
Output: Timestamped log of what the customer clicked, what the UI showed, what they said, ready to drop into Linear or Jira.
Triage stops requiring a human to watch the video at 1.5×. The ticket arrives pre-written.
Education
Lecture archive with visual anchors
Who: course creator publishing a 60-video curriculum.
Input: recorded lectures, ultra-detailed (8×) for whiteboard segments, regular (2×) for intros.
Output: transcripts where every diagram, code snippet, and slide change is described inline with the spoken explanation.
Students search text, not video scrubbers. A diagram at 14:32 becomes findable by what it shows.
Sales / RevOps
Sales calls with on-screen numbers
Who: RevOps lead processing Gong or Zoom recordings of demos and discovery.
Input: 45-minute recording where the prospect shared a dashboard.
Output: transcript merged with frame captures of the dashboard, pricing screen, and competitor comparison the prospect flashed at 31:08.
CRM notes capture what was said. vid2md also captures what was shown.
For builders
One API. Three calls.
POST a video URL → poll for status → GET the .md.
curl https://vid2md.com/v1/jobs \
-H "Authorization: Bearer v2m_..." \
-H "Content-Type: application/json" \
-d '{"video_url":"https://...","mode":"concise"}'