| name | ozor-url-to-video |
| description | Turn any URL — landing page, blog post, product page, documentation page, changelog, public PDF, or web article — into a finished AI-generated video using the Ozor MCP. Use this skill whenever the user gives a link and wants a video from its content: 'make a video from this URL', 'video from my landing page', 'turn this blog post into a video', 'video from https://...', 'video from our homepage', 'generate a video from this product page', 'create a video from this article'. This skill runs analyze_document with the url parameter (no scraping needed on our side), presents the resulting plan to the user, then generates → exports → embeds. Do NOT use this skill for a short text brief (use ozor-generate) or for an attached local file (use ozor-document-video). |
Ozor URL-to-Video
End-to-end web-page-to-video on Ozor.ai. Hand Ozor a URL, get a plan back that reflects the actual page content (headlines, value props, social proof, CTA), approve or edit, and ship a finished video.
When to use
Use this skill when the input is a URL. Signals:
- "Make a video from https://..."
- "Video from my landing page / homepage / product page"
- "Turn this blog post into a video"
- "Video from this article / changelog page"
- Any message that contains a URL the user expects you to read
Do NOT use this skill when:
- Input is a short text brief →
ozor-generate
- Input is an attached local file (PDF/PPTX/DOCX) →
ozor-document-video
- User wants a written prompt only →
ozor-landing-page-to-video (prompt-only skill)
- User wants to manage existing videos →
ozor-video-library
Required MCP tools
mcp__ozor__analyze_document — accepts a url parameter directly; no upload needed
mcp__ozor__get_plan / mcp__ozor__update_plan — review/edit the plan before generation
mcp__ozor__generate_from_plan — kicks off the build, returns projectId
mcp__ozor__wait_for_job — if the response includes a jobId
mcp__ozor__export_video, mcp__ozor__wait_for_export
mcp__ozor__get_embed_code
If the MCP isn't connected, tell the user and stop. Do not browse the URL yourself and hand-write a prompt — Ozor extracts brand colors, hero imagery, and structure directly from the page. Your scraped text would lose that.
Workflow
Step 1 — Hand the URL to Ozor
mcp__ozor__analyze_document({
url: "<the URL>",
...optional hints: targetDuration, format, audience, tone, focus
})
For URLs:
- Landing pages / product pages → default to
format: "9:16" or "16:9" (ask if not obvious — vertical is great for social, landscape for embedding on the site itself), targetDuration: "30s"–"45s", tone "confident".
- Blog posts / articles →
format: "16:9", targetDuration: "60s"–"90s", tone "informative".
- Changelog / release pages →
format: "16:9", targetDuration: "30s"–"45s", tone "energetic", focus "one scene per shipped feature".
- Documentation pages →
format: "16:9", targetDuration: "60s", tone "clear, technical", focus "explain what this feature does and how to use it".
Returns { planId, plan }.
Step 2 — Present the plan and STOP
Render the plan to the user concisely:
📋 Ozor plan for <URL>
**Source:** <page title or domain>
**Format:** 16:9 / 9:16
**Estimated duration:** ~X seconds
**Detected brand:** <colors / product name if Ozor surfaced them>
**Scenes:** N
(1) <scene title> — <one-line direction>
(2) <scene title> — <one-line direction>
...
Approve as-is, or change anything? You can edit scenes, swap format, change tone, or shift focus.
Wait for approval. Don't call generate_from_plan yet.
Step 3 — Apply edits (if any)
Use mcp__ozor__update_plan({ planId, ... }). Re-show the plan with get_plan. Loop until approved.
Step 4 — Generate
mcp__ozor__generate_from_plan({ planId }) // -> { projectId, jobId? }
If a jobId comes back, mcp__ozor__wait_for_job({ videoId: projectId, jobId }).
Step 5 — Export
mcp__ozor__export_video({ videoId: projectId })
mcp__ozor__wait_for_export({ videoId: projectId }) // -> shareUrl, editorUrl, downloadUrl
Step 6 — Embed
mcp__ozor__get_embed_code({ videoId: projectId })
Step 7 — Present results
🎬 Video ready from <URL>
**Share:** <shareUrl>
**Editor:** <editorUrl>
**Download:** <downloadUrl>
**Embed (drop this on your site):**
```html
<iframe ...></iframe>
Want a vertical 9:16 cut for social, or a longer 90s explainer version? I can regenerate from the same plan.
## What the URL flow does best
- **Landing pages** — Ozor picks up the hero headline, the value prop, social proof, the CTA, and brand colors. Your prompt would lose all of that.
- **Blog posts** — picks up the thesis and section structure; produces an explainer that maps to the article.
- **Changelogs** — naturally maps each release/feature to its own scene.
- **Product pages** — pulls product screenshots / hero imagery into scenes when available.
## Format-decision heuristic
If the user didn't say:
- URL is a homepage or product page **and** the obvious use case is social/ads → `9:16`
- URL is a homepage or product page **and** the obvious use case is embedding on the page itself → `16:9`
- URL is a blog / article / doc / changelog → `16:9`
When ambiguous, propose `16:9` as default and offer a `9:16` rerun at the end.
## Variants worth offering after the first render
Mention these only after the user has the first video — don't preemptively generate them:
- **9:16 social cut** of a landscape video → regenerate from the same plan with `update_plan({ planId, format: "9:16" })`.
- **Short teaser (15s)** + **full explainer (60s)** from the same page.
- **Two audiences** (e.g. "technical buyer" vs "executive buyer") — re-analyze with a different `audience` hint.
## Errors and recovery
- **`analyze_document` returns a thin plan** (often: paywalled page, JS-rendered SPA Ozor couldn't read) → tell the user, ask for an alternative URL or whether they can paste the key content as a brief for `ozor-generate`.
- **Detected brand colors look wrong** → `update_plan` with corrected colors before generating.
- **Generation fails** → check the scene the error references, fix via `update_plan`, retry.
- **Export hangs** → call `wait_for_export` again; it resumes.
## Rules
1. **Always pass the URL to `analyze_document`.** Never scrape the page yourself and route to `generate_video` — that throws away Ozor's brand and visual extraction.
2. **Always show the plan and wait for approval** before `generate_from_plan`. The plan is where format and tone get locked in.
3. **Use `projectId` as `videoId`** for export, wait_for_export, and embed.
4. **Surface every URL** — share, editor, download, embed.
5. **Propose, don't pre-generate, variants.** Don't burn renders on 9:16 + 16:9 + teaser without being asked.