Opus 5.5 Can One-Shot a Launch Video for a Few Dollars in Tokens — LaunchVideo Shows the Agency Business Hiding Inside That Number.

LaunchVideo did not launch a video generator.

It launched a receipt.

The Show HN post lays out exactly how the pipeline works: Claude Opus 5.5 "writes the film" as code — not by calling a video generation model, but by writing HTML that gets rendered deterministically into video frames. The cost is "roughly 90k input and 15k output tokens per film, most of it the HTML itself," and the whole thing takes "about four minutes" to produce.

That is a business plan disguised as a technical writeup.

What's actually happening under the hood

The pipeline runs three tools inside "one session in a fresh microVM" that gets "thrown away after":

• web_fetch — pulls facts and copy from the product's URL or the text prompt
• check_scene — validates that a rendered frame actually looks right before moving on
• render_video — compiles the final output

Output specs: "1920x1080 at 30 fps," compressed at "crf 18," rendered on "Amazon Linux 2023 on arm64, 4 vCPU, 8 GB RAM, Node 22" using headless Chromium and static ffmpeg, with a virtual clock replacing all timing mechanisms so the render is deterministic — same input, same output, every time.

Every example on the page is "a URL or a prompt in, an MP4 out. No edits."

The commenters already told you the price

HN thread commenters weren't speculating about cost — they'd already run it. One pointed to a project made "entirely with Opus 5.5 + $3.21 of OpenRouter API usage." Another reported Claude Code produced comparable work for "about $4." One founder, syrusakbary, described the before/after directly: "I spent 6 hours creating a video for our Wasmer SDK announcement... then yesterday I tried Opus 5.5... and man, I'm impressed. The video was one-shotted."

Six hours of manual editing versus a one-shot render for a few dollars. That gap is the entire opportunity — someone still has to know the pipeline exists, run it well, and stand behind the output.

Where the real objection is, and how to price against it

Not every comment was glowing. Several called the output "flashy slide decks" and "low effort, low quality content," and one was specific about why: "there are no drawings and animations that explain the actual concepts."

That's not a reason to skip the business. It's a pricing tier. A base tier sells the fast, cheap, one-shot render for founders who just need something to post today. A premium tier sells what the AI-only version skips — real diagrams, custom animation, and concept-level explanation for technical products where a slide deck with transitions doesn't actually teach anything.

The moneyPlay in practice

1. Build the same prompt-to-code-to-render pipeline (or wrap LaunchVideo directly) and offer a 24-hour turnaround on launch videos for indie founders

2. Quote a flat per-video price that already assumes ~$3-4 in token cost and a few minutes of render time — your margin is speed and reliability, not compute

3. Add a QA pass using LaunchVideo's own check_scene-style validation so you never hand a client a broken render

4. Upsell a "concept explainer" tier with real diagrams and motion graphics for products that need more than a slide deck

5. Turn one-off clients into a retainer — every future release gets a video without a new creative brief

Bottom line

The technical writeup already published the unit economics: ~100k tokens, four minutes, a few dollars. The agency opportunity isn't discovering that price — it's being the one who packages it, guarantees the output, and answers the "these are just flashy slide decks" objection before the client raises it.

Sources:

https://news.ycombinator.com/item?id=49836374

https://launchvideo.io