Skip to content
ScreenCI

Who Does What

ScreenCI is one video library for the whole SaaS team. Everyone uses the same buttons and the same browser editor; the difference is which buttons you reach for and what you never have to do again. This page walks through the three roles that make most product videos. Every button below is described in Make videos by prompt, and the Prompt cookbook has example descriptions for each.

You will learn

Product marketing

Goal: the launch video ships the same day as the feature, in your branding, in every market’s language, at a link you embed once.

What you do:

  • Add project or Add video with one sentence about the feature. The agent records your real app, and the tab opens the video.
  • Edit in the browser: rewrite a narration line, swap the voice, move an overlay, cut a slow section. Nothing to prompt for.
  • Add a language from the language menu: pick the language, paste the prompt, and the localized version records on the localized page. See Languages.
  • Export the finished version and share its public URL. The link always serves the version you selected, so an update never means a new embed.

What you never do: schedule a recording session, wait for an engineer to be free, re-export the same video in six languages by hand.

Docs and support

Goal: help-center walkthroughs and docs screenshots that stay correct after every release.

What you do:

  • Add video for each flow customers ask about, in the words they use: “how to invite a teammate and set their role”.
  • Add screenshot for docs figures and README images. A still from the same script, branded and cropped, re-captured whenever the UI changes. See Screenshots.
  • Edit narration and subtitles in the browser when the wording changes.
  • Re-record a video after the product changed under it: one click when CI records the project, a prompt otherwise. The script is only touched where the product changed.

What you never do: notice a stale video from a customer ticket. Once an engineer has done Add to CI, the pipeline re-records the videos on every push, and a flow that broke fails the run.

Engineering

Goal: own zero video tooling, and never record a video yourself.

What you do:

  • Run one prompt inside the repository once. The agent commits the screenci/ scripts to the product repository and opens a pull request. From then on the videos live with the code and every one is a Playwright E2E test.
  • Add to CI once. The agent stores a project-scoped key in your CI provider, adds a pipeline that records on every push, and triggers the first run. Any provider works.
  • Review the pull requests the agent opens when a teammate’s Edit or Add video prompt ran inside the repository. See Repository and CI.
  • When you prefer to start in the repository yourself, paste the integration prompt into the agent there, or write the scripts by hand.

What you never do: record, edit, or localize a video. A stale video is a failed CI run, which you fix like any other broken test: click Edit on the video and describe the change, or fix the script.

The buttons by role

Button Marketing Docs and support Engineering
Add project yes yes yes
Add video yes yes yes
Add screenshot yes yes
Edit (prompt) yes yes yes
Edit in the browser yes yes
Add a language yes yes
Record all, Re-record yes yes yes
Add to CI yes
Export, public URL yes yes

Every button is available to every member; the table shows who usually clicks it. Set AI context and Branding once and every prompt in the organisation starts from them.

What’s next