Skip to content
HTML/CSS to ImageDocs

Record HTML/CSS or a live webpage as an MP4, WebM, or animated GIF. Recordings use the same Chrome renderer as images, so your fonts, CSS, and JavaScript work the same way.

Use them for scrolling website walkthroughs, product demos, animated social posts, and CSS or JavaScript animations. Recordings have no audio.

Send html or url, a duration_ms, and optionally some actions to POST /v1/recording:

Terminal window
curl -X POST 'https://hcti.io/v1/recording' \
-u "$HCTI_API_ID:$HCTI_API_KEY" \
-H 'Content-Type: application/json' \
-d '{
"url": "https://example.com/",
"viewport_width": 1280,
"viewport_height": 720,
"duration_ms": 10000,
"actions": [{ "type": "simple_scroll" }]
}'
{
"id": "01a1b2c3-d4e5-7f60-8a9b-0c1d2e3f4a5b",
"recording_url": "https://hcti.io/v1/recording/01a1b2c3-d4e5-7f60-8a9b-0c1d2e3f4a5b.mp4",
"status_url": "https://hcti.io/v1/recordings/01a1b2c3-d4e5-7f60-8a9b-0c1d2e3f4a5b/status"
}

Use recording_url like an image URL. Download it, or put it straight into your website, app, or email:

<video src="https://hcti.io/v1/recording/01a1b2c3-d4e5-7f60-8a9b-0c1d2e3f4a5b.mp4" autoplay muted loop playsinline></video>

If the recording isn’t finished when the URL is requested, the request waits and then returns it. For background jobs, you can check its status instead.

See Using the API for the full request and response.

Parameter What it does
duration_ms Required. How long the recording is, from 500 ms to 30 seconds (90 in target mode).
actions What happens while recording: simple_scroll down a page, scroll to exact spots, click, stall, execute_javascript, and branches for elements that may not exist. Leave it out to record the page as it is.
duration_mode How the actions fit into duration_ms. at_most (default) ends when the actions finish. fixed is always exactly duration_ms. at_least lets the actions run long. target speeds up or slows down the video to fit.
trigger When recording starts: right away, when an element appears, or when your page calls window.RecordingReady().
output Format (mp4, webm, gif), frame rate, and scale. Defaults to MP4 at 30 fps.
viewport_width / viewport_height The video size in CSS pixels. HTML recordings capture the whole viewport, so set these to the size you want.
device_scale Pixel density. 2 gives sharp, retina-quality text.
defer_render Wait to render until the recording is first requested.

Many image parameters work too, like css, google_fonts, block_consent_banners, proxy_id, and storage_destination_id. The parameters reference marks which ones work with recordings.

Recordings use the same monthly allowance as images. A recording’s cost depends on its requested duration_ms and duration_mode. Format, frame rate, and resolution don’t change it.

  • 2 credits for the first 5 seconds, plus 1 credit for each additional 5 seconds or part of 5 seconds.
  • target and at_least use twice as many, because the capture can run longer than duration_ms.
duration_ms at_most / fixed target / at_least
500–5,000 2 4
5,001–10,000 3 6
10,001–15,000 4 8
15,001–20,000 5 10
20,001–25,000 6 12
25,001–30,000 7 14
60,000 (target only) — 26
90,000 (target only) — 38

For example:

  • A 3-second GIF in fixed mode uses 2 credits.
  • A 10-second scroll in the default at_most mode uses 3 credits, even if the page reaches the bottom after 7 seconds.
  • A 12-second demo uses 4 credits in fixed mode, or 8 in at_least mode.
  • A 30-second target video uses 14 credits, however long the capture was.

Credits are used when the recording is created. Automatic retries don’t use more. A request that matches an existing recording through dedupe_duration_s uses 0 credits. The x-renders-consumed response header shows what each request used.

Manage existing recordings with the Recordings API, or ask an AI assistant connected to the MCP server to create them.

Need help?

Talk to a human. Email support@htmlcsstoimage.com and we’ll help you get started.