Let the AI record your product with Record with AI

Record with AI films your product for you. You describe the steps, and the AI opens your site or app in a browser in Clevera's cloud, works through them in one recorded pass and hands you a screen recording for your video. You can watch it live, send it notes, and take the browser over whenever you need to.

Use it when you don't have time to record yourself, don't have the desktop app to hand, or want a clean, steady walkthrough of a web app.

Before you start#

• You need a web address the AI can open: a public https:// page. Desktop apps and pages on your own computer or company network can't be recorded this way. Use the desktop app for those.
• If the steps happen behind a sign-in, decide how the AI gets in. See Sign in or take over during Record with AI.
• Use a demo or test account where you can. Everything on screen ends up in the recording.
• A role that can create. Editors can't start Record with AI: the option is turned off with "Your role can edit existing content but cannot create new content."
• Only the person who started a recording, the workspace owner and admins can open its live page, send notes, take over or stop it.
• Record with AI uses credits for how long the AI works. See what it costs.

Start a recording#

1. Open Record with AI#

   On the Compose page of any video type, click + Add in the brief box and choose Record with AI in the With AI group. Its tip reads "Let AI record your site or app in the cloud."

   The Record automatically with AI dialog opens: "AI opens your site in a cloud browser and records the steps you describe."

   

2. Describe what to show#

   In What should AI show?, write the steps in order, as you would to a colleague. For example: "Open Projects, create a project called Q3 Launch, then invite [email protected] as an editor." You have up to 2,000 characters.

   The AI does only these steps, in this order, and then stops. It doesn't explore other pages or menus on its own.

3. Add the start page#

   In Start page, enter the first page the AI should open, for example https://app.acme.io/projects. "Public https:// pages only." Starting on the right page saves time and credits.

4. Choose how the AI signs in#

   Under Sign-in, pick one:

   Option	What happens
   None	"The page opens without signing in." For public pages.
   Use a login	"AI signs in with an account you give it." Enter Email or username and Password, and tick "I'm allowed to give Clevera access to this account."
   I'll type it myself	"AI stops at the sign-in; open Watch and sign in there." The browser is handed to you at the sign-in page.

   If your workspace has a saved sign-in for this site, the dialog says "Using your saved sign-in for app.acme.io." and the AI starts signed in. Details in Sign in or take over during Record with AI.

   Whenever the recording signs in, Don't include signing in in the video appears, ticked: "The video starts once you're signed in." Untick it if you want the sign-in to be part of the video.

5. Decide whether the AI may change things#

   Allow changes ("Let AI create or edit things in your app") is off by default. Turn it on when your steps create or edit something, such as a project, a record or a setting.

   While it's off, the AI only looks around: it may open, view, scroll, search, filter, sort, switch tabs and sign in, but it doesn't create, edit, delete, send, publish, invite, buy, pay or change a setting, and it doesn't confirm a dialog that would. If one of your steps needs a change, the AI shows where it's done, for example by opening the form or menu, then stops before saving or submitting and goes on with your next step.

   When your steps change something and Allow changes is off, the live page says so as soon as the recording starts: "Some steps change things and Allow changes is off. AI will pause there and ask you." If the AI hands you the browser at that step, click Allow changes and continue to let it make the change.

   

6. Start recording#

   Click Start recording. A tile appears under your brief with the status Recording… and a Watch button.

How the recording is made#

The AI works through your steps once, in one browser, and that pass is your recording. Before it starts, it turns your steps into a short plan: a goal and numbered steps, which you can open on the live page. Clevera then shapes the film so it plays like a confident demo:

• It starts at your first step. Opening the browser and waiting for the site are left out, and so is signing in while Don't include signing in in the video is ticked.
• The AI's thinking time is cut out. While the AI reads a page or decides what to do next, the screen stands still. Clevera removes those still stretches and keeps a short moment before each action and a readable pause after it, so you see every click, scroll and result. The film opens on your first screen and ends shortly after the last result appears, without a long hold on it.
• The AI acts directly. It goes straight to a page whose address it knows, scrolls through a page when a step asks to show it, and waits a moment after each change so the result is on screen.
• Typing shows on screen. The AI types into fields letter by letter, so viewers can follow what goes into each one, and a filled form stays on screen for a moment before it's sent.
• Pop-ups and ads are cleared away. The AI closes cookie and consent banners, sign-up pop-ups, chat bubbles and covering ads as soon as they appear, always with the most private choice (such as rejecting tracking). It never clicks an ad. An ad or consent screen that covers the page and that the AI didn't open is cut out of the film.
• It fixes its own slips. If a click missed or a form wasn't sent, it tries again another way, and the activity says "That step did not go through. Trying it again."
• A stuck step is skipped. Each step gets twice its even share of the working time, between 1.5 and 4 minutes. If the AI is still on a step after that and hasn't made progress in the last 45 seconds, Clevera moves it on to your next step in the same browser. A step that's still moving gets until 45 seconds after its latest action, and never more than one and a half times its time. The activity then says, for example, "Step 3 of 5 took too long and was skipped." If that happens on the last step, the recording ends there: "Step 5 of 5 took too long, so the recording ends there." What was filmed up to then is kept.

Because the still stretches are cut, the recording jumps a little where the AI paused. Your video is shorter than the time the AI spent working.

Watch it live#

Click Watch on the recording's tile to open the live page. The browser fills the left side. The panel on the right is headed with the AI's goal (or "Recording app.acme.io" before it has one); hover the ⓘ beside it to see "What you asked for".

Above the browser, one line says whether what you see is being filmed:

On the right, a card shows the stage, Getting ready → Recording → Finishing, with one headline and one line of why:

Under the headline, a line shows the step the AI is on, such as "Step 2 of 3: Create a project called Q3 Launch". Click Steps to open The AI's plan: the goal, your steps numbered with a check for each one done and "Doing this now" on the current one, and what the AI Keeps in mind. The plan is made from your request.

Below the card, Activity lists what the AI did, grouped under the step it served. Long steps fold their earlier lines behind Show n earlier. If you scroll up, a New activity button takes you back down when something new arrives.

You don't have to watch. Click Back, keep working on your brief, or go elsewhere in Clevera: "The AI keeps going. You can come back to this page." While the recording runs, a Your AI recording button stays in the corner of other pages. If the AI needs you, it chimes, a dot appears on the tab and the button says "We need one quick answer".

Send the AI a note#

While the AI works, type a note in the box at the bottom right, for example "Try the Reports tab next", and click Send note (or press Enter). "The AI reads it before its next step and adjusts." Notes are up to 280 characters each, and they win over your original steps.

Take over or stop#

• Take over gives you the browser, to type something private, fix a wrong turn or set up the page: "Use the browser yourself. The AI waits until you hand it back." Click Hand back to the AI when you're finished. See Sign in or take over during Record with AI.
• Stop ends the recording. Click Stop, then Stop session at "Stop this session?" ("The AI stops and the session ends. You can't pick it up again."), or Keep going to carry on. There's no pause button: to make the AI wait, take over. A recording you stop is charged for the time the AI worked before you stopped it.

When the recording is ready, click Back to Compose to return to your brief.

What you get#

When the recording is ready, its tile in Compose says Ready, and your video uses it like any screen recording:

• How-to, product demo, polish and article videos use it as the recording the video is made from.
• Social clips use it as product footage, alongside your pictures and links. It counts as one of the 3 videos a clip takes.

The recording is named with a short title the AI writes from your steps, 3–8 words in the language you wrote them in, rather than the steps themselves. You see it on the tile and in your recordings.

The cursor is redrawn smoothly from the AI's real clicks, so zooms and click effects work as they do on a recording you make yourself. What you do while you have the browser stays in the recording too, except signing in while Don't include signing in in the video is ticked.

The recording is also saved with your workspace's recordings, so you can use it again.

Using it from the setup chat#

The setup chat can suggest recording your product. It shows a card with Start page, What it should show, Sign-in (No sign-in or I'll sign in myself) and an It may create or edit things box, then Start recording. The box works like Allow changes: it's unticked by default, and the AI then stays read-only. With I'll sign in myself, the card also has Don't include signing in in the video, ticked by default. The live browser opens above the chat, with Take over and Stop (confirm with Yes). "At the sign-in page the browser becomes yours. Sign in there, never in the chat." To give the AI a login instead, use Record with AI from Compose.

Limits#

Time you spend holding the browser doesn't use up the AI's working time. After you hand it back, the AI always gets at least two more minutes.

When it doesn't finish#

On Compose, a recording that didn't finish shows Didn't finish with the reason, and Generate says "The recording of app.acme.io didn't finish." Click Try again to reopen the form with your steps filled in, Remove to take it off, or See what happened to open the live page.

If a recording fails on our side and can't be recovered, every credit it used goes back to the kind of credits it came from, and the Usage page shows a "Failed run refunded · AI screen recording" line. A recording you stopped stays charged for the time the AI worked.

A recording where a step was skipped still finishes and is Ready. Watch it before you generate: if a skipped step matters, record that step on its own with clearer instructions, or take over and do it yourself next time.

Messages when you start#

If the session is interrupted#

Sometimes a recording is cut short by a problem on our side, for example when the cloud browser or the server running the recording stops partway through. The cloud browser films everything it shows, so Clevera can usually still get that footage back.

A recovered recording works like any other and is cut the same way: it starts at your first step, leaves out a sign-in you asked to leave out, and drops the AI's thinking time. It holds only what was filmed before the interruption. The AI doesn't go on with steps it hadn't reached, so watch it before you generate. If the last steps are missing, start a new recording with your steps, or with only the missing ones. Either way, the recovered recording stays in your library. If you had taken the browser and never handed it back, that stretch is cut out.

A recovered recording is billed like a finished one, because you get its footage. While Clevera is still trying to recover it, the credits stay charged; if it can't get the footage back, the run counts as failed on our side and its credits are returned.

Only interruptions on our side are recovered. A recording you stopped, one that never got past the sign-in, or one where the AI filmed nothing ends with Didn't finish as usual. If Clevera can't get the footage back, the live page says "No recording to watch" ("This session did not leave a video. You can still go back.") and the tile shows Didn't finish. Click Try again to record it again with your steps filled in.

Use a recording again#

Every finished Record with AI recording, including a recovered one, is saved with your workspace's recordings. To make another video from it, you don't need the AI to film the same steps again:

1. On the Compose page, click + Add and choose Choose from library ("Pick a video you've already recorded or uploaded.").
2. Set the source filter to AI browser ("Recorded for you by AI"), or search for the recording by name.
3. Click the recording to preview it, then click Use this file.

Picking an earlier recording starts no new session in the cloud browser, so it adds no AI recording time. Making the video costs the same as from any recording. Anyone in your workspace can pick it. In Content → Recordings, it's marked Recorded for you in an AI browser. See Find and reuse your recordings.

Related#