How to use Eizeno

Creating videos with Eizeno requires Google Chrome on a computer and the Eizeno Browser extension. Production features are unavailable on phones and tablets. Open eizeno.com in Chrome on your computer.

Eizeno turns the URL of a SaaS product into an introduction video. An AI (Claude Opus 5.5) operates the real screens in a browser, records the main features, and edits them into a video (MP4) with narration and music (subtitles if you ask for them) in the taste you choose. You can then refine the video by chatting. This page covers everything from getting started to what to do when something goes wrong.

The overall flow

1. Sign in (with an email link or Google)
2. Top up production tokens
3. Send the URL of the product you want to introduce in the chat
4. With the Chrome extension installed, request a video in chat. The AI automatically opens a dedicated tab to explore and record, and asks you to sign in if needed
5. When the recording is done, pick one of the 20 tastes with "Choose a taste" in the chat (three are recommended), or describe what you want
6. The recording is edited into a video with narration and music (subtitles only if you ask)
7. Tell it in the chat what to change (every version is kept)
8. Download the MP4

Getting started (sign in)

• On the sign-in page, sign in with a link sent to your email address or with your Google account. There is no password.
• The email link is valid for 15 minutes. Open it and press "Sign in" to finish signing in.
• The same email address is the same account, whether you use the email link or Google.
• If the email doesn't arrive, check your spam folder and the spelling of the address. Wait a little before requesting another link.
• By signing in, you agree to the Terms of Service and the Privacy Policy.
• Switch the display language (日本語 / English) with the language button in the left column or under "Settings & top-up".

Top-up and balance

• Your balance is 0 right after you sign in. Answer every question in the profile in the Account section of Settings (role, company, full name, how you heard about Eizeno, what you hoped to get, and the problem you'd like AI to solve) and save it to receive 300,000 free production tokens once per person, enough for about one video. The banner at the top of the screen takes you to the profile. You can edit the profile later, but the tokens are granted only the first time (valid for 6 months). Usage varies with content and revisions; a completed video is not guaranteed. The campaign may end without prior notice. Tokens already granted remain usable until their original expiry date.

• Exploring and editing use production tokens. Open your balance and top-up from the balance in the left column or from "Settings & top-up".
• ¥3,000 buys 1,351,350, ¥10,000 buys 4,504,500 and ¥30,000 buys 13,513,500 production tokens. You can also top up any whole-yen amount from ¥3,000 to ¥999,999 (yen and dollar amounts are whole numbers only; decimals are not accepted). Prices are in Japanese yen.
• A top-up gives you the tokens that spend 50% of its pre-tax price on AI usage. Yen is converted to US dollars at a fixed rate of ¥180 per US$1, and US$1 of AI usage equals 162,162 production tokens (fractions rounded down). The rate does not follow daily exchange rates. The top-up screen and the checkout page show the tokens for the amount you choose.
• Amounts are shown before tax; tax is added at checkout (with Japan’s 10% consumption tax, a ¥3,000 top-up costs ¥3,300). Pay by credit card on the Stripe checkout page.
• Customers outside Japan top up in US dollars (overseas pricing): $20 buys 1,621,620, $70 buys 5,675,670 and $200 buys 16,216,200 production tokens (50% of the price goes to AI usage), or top up any whole-dollar amount from $20 to $5,000. When you use Eizeno from outside Japan, top-ups are in US dollars; from Japan, they are in yen (the currency can't be chosen).
• US dollar payments require a billing address outside Japan and a card issued outside Japan (customers in Japan pay in yen with consumption tax). Payments that don't meet these conditions, or whose country can't be confirmed, are refunded in full automatically and no production tokens are added (the payment shows as refunded in "Payment history", and we let you know by email).
• No tax is added to US dollar payments. Validity (6 months from purchase) and usage are the same as for yen.
• Tokens are valid for 6 months from purchase, and the earliest-expiring tokens are used first. "Purchases and expiry" shows what remains of each purchase and when it expires.
• There is no subscription. Only what exploring and editing actually use is deducted.
• "Usage by exploration" shows the production tokens each exploration used.

Starting an exploration

Open "New video" in the left column and send the URL of the product in the chat. You can add the features you want to show.

Example: https://example.com Focus on pricing and the free-trial flow

URLs that work:

• The address of a public site that starts with https:// (example.com without https:// also works)
• Pages that open without signing in (a home page, a pricing page, features you can use without an account, and so on)
• Services that need a sign-in (sign in yourself in the dedicated tab when needed; see Services that need a sign-in)
• Your own product, or one you are authorized to operate and record

URLs that don't work:

• localhost, IP addresses, and internal network addresses (such as .local and .internal)

Tips:

• Mention the features or the order you want, and the exploration focuses on them.
• One exploration runs at a time. Send the next URL after the running exploration or edit has finished.
• Starting an exploration needs a minimum balance. If yours is short, the screen shows the amount needed and a link to top up.

Use your Chrome tab (extension)

Videos are made using the Chrome extension. Install it, then send the product URL and your video request in chat. It connects automatically and opens a dedicated tab for exploration and recording. No connection code or extension start button is needed. If the extension is missing, installation instructions appear in chat. Your production request starts sending dedicated-tab screens, text and action records to Eizeno; screens and text are processed by Anthropic’s AI.

Install the extension

• The extension is available only from the Chrome Web Store. In the extension instructions shown in chat, choose “Install from the Chrome Web Store”, then choose “Add to Chrome” on the store page.
• After adding it, reload your Eizeno production page. The extension connects automatically.
• If you previously loaded the extension manually from a ZIP, remove that copy at chrome://extensions, then add it again from the store.

Update the extension

Chrome updates extensions added from the store automatically. From 1.10.0, the extension also checks for a newer version each time an exploration starts from the production chat, and updates itself before exploring if one is available (no need to reload this page). It never updates during an exploration. To update right away, or if it could not update automatically, enter chrome://extensions in Chrome’s address bar and enable Eizeno Browser. Turn on “Developer mode” in the top right and click “Update”. After updating, restore your previous Developer mode setting and reload this production page. If the update has not reached you yet, wait a while and try again.

Connect a tab and create a video

• Send the product URL and your video request in chat. The extension automatically connects and opens the target URL in a dedicated new tab. Existing tabs are not operated.
• If a link or a redirect takes the tab to another website during exploration, the exploration does not stop. Pages related to the product (sign-in, help or app pages) are explored; if the site is clearly unrelated, the AI goes back with the browser's back. Sites other than the product are opened as public pages without cookies (except while you sign in yourself). If the tab lands on an error page or an address that cannot be opened, it goes back to the previous page, and the exploration stops only when it cannot go back. During sign-in, you can continue even if the sign-in moves to another domain (such as a dedicated sign-in site).
• The extension popup is for checking status, stopping, or resuming after manual sign-in. You do not need to open it to start production.
• The dedicated tab appears in the “Eizeno is exploring” group. Exploration can usually start while you work in another tab. If the chat asks you to open the dedicated tab, select it once. Exploration and recording then start automatically.
• If the extension cannot connect after an update, reload the page for this run.
• If the extension’s connection expires, check that Eizeno Browser is enabled in Chrome and ask to explore again in the same chat.
• If sign-in is needed, AI control and recording pause and the chat asks for consent (use the buttons, or answer in the chat, for example "yes" or "continue without signing in"). After agreeing, sign in yourself in the dedicated tab, then send "I signed in" in the chat (the extension's Finish sign-in button also resumes). Sign-in screens and inputs are not sent or recorded. If you decline, exploration continues on public pages.
• If Chrome is still signed in to the product and the dedicated tab opens a signed-in screen (at the start, or later during exploration), that screen is neither recorded nor sent to the AI. Exploration pauses and the chat asks whether the signed-in screens may be explored and recorded. If you agree, exploration continues view-only. If you decline, exploration stops without recording (if it happened mid-way, exploration returns to the previous public page and continues). To show only public pages, sign out of the product in Chrome, or send the URL of a public page that opens while signed in. Eizeno's own landing page is available at https://eizeno.com/lp.html.
• Stop in the extension switches the dedicated tab to a stopped screen to discard delayed page actions. Choose "Return to product" to browse yourself. The tab stays open after completion; close it when no longer needed.
• After recording, return to Eizeno, choose a taste, and edit and download the introduction video as usual.

Scope of control

• The extension operates only the dedicated tab it opens for production. It does not export cookies or saved passwords to Eizeno. During exploration, screen images, page text and interaction records are sent to Eizeno to produce the video.
• The AI views the product only; it does not purchase, delete, send or sign in. Connect only services you are authorized to operate and record. Use a demo account.
• Exploration stops on unsupported pages or when protections cannot be installed. It does not switch to another browser mode.
• The extension uses your own Chrome network connection, so it cannot fully prevent access to internal-network destinations. Do not use it for internal-network services.

Watching progress and stopping

If exploration stops, the chat and canvas show the reason and guidance for retrying. If the reason is unknown, we say so.

• The canvas on the right shows the screen the AI is operating, live.
• While the introduction video is being edited, the canvas shows the current step (planning, narration, music, choosing the film styles to learn from, the storyboard, rendering, mixing, saving, thumbnail, and so on), the elapsed time, and the time since the last update. Only the rendering step also shows its measured progress (%). No overall time remaining is shown. Editing continues on the server even if you close or reload the page, and the same state is shown when you reopen it.
• If there is no update for a while or the connection is lost, the canvas says so and shows what you can do next, such as reloading or stopping. If the editing job stops responding for a long time, it is treated as interrupted. Send your request again to remake it.
• The chat lists the features found and the AI's notes as it goes, along with the production tokens used.
• The stop button (■) in the input box stops an exploration, a taste proposal or an edit at any time. Writing "stop" in the chat also stops it. Only the tokens used up to that point are deducted (including requests already sent to the AI before you stopped), and when an exploration is stopped, the recording so far is kept. Stopping a taste proposal keeps the recording and any earlier taste options, and you can ask for new ones.
• An edit that is interrupted partway is not treated as finished. Earlier versions are kept, and sending the same request again retries it.
• It also stops when your balance runs out or the time or step limit is reached. The recording so far is kept.
• If an exploration stops early but the recording so far and the features it found are kept, choose "Make it from the recording so far" in the chat, or ask in the chat to make the video from the recording so far, to continue with the taste proposal. Nothing starts automatically when it stops (the taste proposal and later steps use production tokens). If no recording is kept, no feature was found, or the recording is not used for safety, ask in the chat to explore again.
• Reopen past explorations from "Past explorations" in the left column.

Choosing a taste or describing your own style

• When the recording is done, the chat says "Pick a taste from the list, or describe what you want in detail in the chat" with a "Choose a taste" button. The button opens a list of 20 tastes. The canvas step "Choose a taste or describe it" shows only the status and the taste you chose (or what you wrote).
• A taste is a template that sets the layout, the typeface and how text appears, the motion, the background, section cards, pacing, voice and music (keynote, live demo sale, documentary, release notes, training steps, newscast, before and after, and more). In every taste the real screen is shown inside a browser window with its real address and a "real screen" label, and the camera only moves in at moments of interaction.
• Each taste in the list has a name, a one-line description, the uses and mood it suits, and a short sample made with a fictional service. The sample is still at first and plays muted when you hover over it; with reduced motion turned on, only a still image is shown. Your finished video, colours and audio will differ.
• The three that suit your product best are marked "Recommended" with the AI's reason and listed first. You can filter by use (overview, announcement, training, social, sales) or mood (calm, energetic, premium, friendly). Close the list with Esc or the close button.
• Press "Use this taste" to start editing the recording into a video right away.
• Colours are taken from the site that was explored (its main colour, background and text colours) and fitted to the taste's colour scheme. If a combination would be hard to read, the same hue is made lighter or darker. Colours or a design you ask for come first.
• The video is for promotion, so it has no subtitles by default. To add them, write something like "with subtitles" (separately from narration on or off). After it is made you can also ask in the chat to add or remove subtitles. Headlines and other on-screen text are separate from subtitles and always shown.
• If you write a request in the chat before choosing, it is taken into account. Wishes you write after choosing a taste (length, voice, colours, features to show and so on) are applied on top of the taste. You can also skip the tastes and start from a written request alone (for example: "Calm mood, no narration, with subtitles"). If your wish changes the presentation itself ("like slides", for example), the video moves away from the taste's template.
• The length of the video is set when it is rendered, for example by the time the narration takes to read. When a version is ready, the chat tells you its actual length (and why, when the narration made it much longer or when a purpose-built video is far from the length you chose).
• Tastes marked as using generated shots place an AI-filmed camera shot before set scenes: a push-in onto a laptop screen on a desk that flows straight into the real screen recording, or atmosphere footage without the product's screen (marked "Illustrative footage"). Every shot is checked frame by frame; broken parts are cut, or the shot is left out (the video is still finished without it). Shots use more production tokens. The samples in the list show placeholder footage instead of real shots.
• On request, an AI-generated opening shot can also be added, which uses more production tokens.
• "Free-form (AI-coded)" (custom, no template) appears right after the recommended tastes. Choosing it means the AI first decides how to stage the requested atmosphere (genre, tempo, how text appears, how lively the motion is) and then designs the layout, typography, colours and motion for this particular video from scratch instead of assembling fixed parts. To do this, it picks the closest film styles from a library of highly regarded product films studied for technique only (shot structure, framing, camera moves, how text appears, timing with sound), and first draws up a storyboard that applies them to this product's recording and materials (other companies' footage, words and logos are never used). It may also use the generated shots above when they suit the staging. The AI decides how the real screen is framed (in a window, as a large crop, tilted...), but never changes its content or colours, and the "Real product screen" label and the real domain stay next to it whenever it is visible. The result is checked automatically (text is readable, nothing runs off the screen, the screen and the site's material are clearly visible, material keeps its proportions and shows its source) and improved before it is finished. If some points still fall short of these checks, the video is still finished in the layout you asked for and you are told; you can refine anything by chat. Only when the rules for showing the real screen can't be kept (the "Real product screen" label and the real domain, an unaltered screen, no words that aren't on the site and so on) is the video finished with a regular template, and you are told. When you skip the tastes and describe what you want in your own words, this is the default. A video made with a template stays on the template when you only change colours, voice, music, subtitles or length, and switches to this way when you ask for a different atmosphere or presentation. It takes longer and uses more production tokens than usual.
• Even if no recommendations could be made, you can pick any of the 20 tastes with "Choose a taste", or write the mood you want and send it to start editing. If you want other recommendations, say so and new ones are proposed.

How scenes are shown (real screen, site material, animation)

A video combines three ways of showing things, scene by scene. The AI chooses them from what the video introduces (how to use a product, a service, a company, ...) and the material on the site; what you ask for comes first.

• Real screen: the screen recorded during exploration, inside a browser window with the real domain and a "Real product screen" label. It zooms in only at recorded moments of interaction. This is the core of how-to videos and of services whose screens are the value.
• Site material: demo videos, photos, diagrams and logos published on the site, taken out of the page and shown large. They get no browser window and no "Real product screen" label; the source domain is shown small instead (so published material is never presented as the live product screen). A material video's own sound is muted by default.
• Animation: the business, strengths or a flow shown with animated text and shapes, using only what the exploration confirmed. Facts, results and numbers are never made up.

• You can write, for example, "Don't use the website screens; centre it on the demo video on the site", "Show how it's operated", "Use this photo" or "No animation". After the video is made you can also ask "Show the demo video large in this scene" or "Turn the company profile into an animation".
• Only material published on the same domain as the explored site (including its subdomains) is fetched. Videos embedded from external services (such as YouTube), files on other sites, streaming video, material that needs a sign-in, and files that are too large are not fetched; the canvas section "Material on the site" shows why. Material from signed-in screens is not used.
• If a wish can't be met (not enough material, low resolution, can't be fetched), the video isn't silently made differently: the chat tells you why and what was done instead.
• Being published on a site doesn't mean material may be used in a video. Versions that use site material say so; please make sure it's your own (your company's) material and may be used. If not, say "don't use the site's material".

The chat assistant

Everything you write in the studio chat is read by an AI (Claude Haiku 5.5). It looks at the state of your production (the screens explored, the features found, what could not be shown, whether it signed in, the tastes, the video versions) and the conversation so far, and decides what to do next.

• Questions ("What did you find?", "Why did it stop?") are answered from the state of your production.
• Clear requests ("The first recommended taste, about 30 seconds", "Make the opening shorter", "Stop", "Sign in and look again") are carried out. For work that uses production tokens, such as a new exploration or an edit, it asks first when your intent is unclear.
• When you point out something the exploration missed (for example "You didn't sign in, did you? You only saw the landing page"), it explains what the previous exploration saw and offers to explore again with a sign-in. If you ask, it explores the same product again (as a new exploration linked to the previous one).
• Don't write passwords in the chat. When asked, sign in yourself in the dedicated Chrome tab.
• The chat assistant cannot change the exploration's rules (view-only, no buying or deleting, operating only the dedicated tab).
• The chat assistant is part of production and uses production tokens: the actual AI cost of each message, which is very small compared with exploring or editing. The support chat is separate and does not use production tokens.

Revising by chat

• Tell it in the chat what to change and the video is edited again (for example: "Make the opening shorter", "Show the pricing feature first", "Use a calmer narration voice").
• Each revision deducts the production tokens that edit used.
• Choosing a different taste rebuilds the video from the structure up. The versions made so far are kept.
• Sending the URL of a different product starts a new exploration. To explore the same product again, just say so. To start again from another page or subdomain of the same site (for example https://example.com/pricing), include that URL; the reply tells you which URL will be explored. A URL on another site becomes a new production for another product.

Videos for a specific viewer and derived videos

You can make short videos tailored to a prospect without recording again. There's no extra form: just write in the chat.

• Write the viewer's industry, role (for example finance staff or executives), the problem they want to solve, the features to show, and the next step you want them to take at the end (for example "Try it free" or "Book a demo") with its URL. Only what you write is used; nothing you leave out is guessed. Customer names and real customer data are not needed.
• Written before the exploration, it makes the exploration put the features and flows that viewer would use first. Written after the recording, it shapes the choice and order of scenes, the headlines and the narration for that viewer.
• From the same recording you can make videos for other viewers (derived videos), for example "Make one for executives from the same recording". The original video and its versions stay as they are, and each derived video has its own versions. Switch between the original and the derived videos above "Finished video and revisions" on the canvas to compare or download them. Name the video you want to change and only that video is revised.
• The next-step URL is shown on the last screen of the video only when you give it (a public address starting with https://). Eizeno does not send or publish the video anywhere.
• Eizeno does not fill in adoption figures, results, reduction rates or customer names. Numbers that aren't in the recording or its summary are left out, and the chat and canvas list them as claims that aren't confirmed. If they are true and you want them shown, confirm them in the chat.
• If a feature you want shown isn't in the recording, it is left out and Eizeno asks whether to explore again. A new exploration starts only after you say yes.
• Each derived video and each revision deducts the production tokens that edit used. The recording and generated assets such as music are reused.

Landing-page, social and other-language videos

From one product demo you can make videos for different uses. Ask in the chat, for example "Make a landscape video for our landing page", "Make a vertical video for social media too" or "Make a Japanese version".

• Landing page (16:9): a landscape video for your website. It states the product's value at the start, shows a few chosen features, and ends with the next step.
• Social (vertical, short): a short vertical video for phones. It opens with the viewer's problem or the value in a few words and shows one to three features. Even in vertical, the product's real screen is shown in a browser window with its real address and the "Real product screen" label, and close-ups keep the control you used in view. No text is placed at the very top and bottom, where social apps put their own labels (their buttons on the right edge may still overlap the screen).
• Another language (Japanese or English): the headlines, narration, subtitles and voice change language. Tell Eizeno in the chat how to read the product name or special terms aloud (for example Kumoita -> "koo-moh-ee-tah"). The reading is used only for the voice; on-screen text and subtitles keep the original spelling.
• It is not a simple crop: the features shown, the opening hook, the amount of text and subtitles, the closing call to action and the length are rebuilt for the use. Features that aren't in the recording and numbers that aren't in the recording or its summary are not added.
• Before making one, the studio confirms what will be made (format, rough length, language, subtitles on or off, narration on or off, readings) and that it uses production tokens, and makes it only after you agree. Eizeno doesn't make every format automatically at once; videos are made one at a time. If you don't choose subtitles or narration, they stay as in the original video.
• Like the original, you can switch to these videos on the canvas to play, download and revise them. Besides the MP4, versions with subtitles offer subtitle files (SRT and WebVTT), and landing-page, social and other-language versions offer a draft post text. The draft only describes what the video shows and has no link. Eizeno never posts to social media.
• For an original or derived video, mark the version you want as complete, then share it from its details page. Vertical videos are shown upright.

Versions and MP4 download

• The finished video appears in "Finished video and revisions" on the canvas. This step opens when a new version is ready.
• Every version of the video is kept (Version 1, Version 2, …). Compare them under "Versions" on the canvas.
• Save any version with "Download MP4" (1920×1080). The exploration's recording can be saved with "Download recording".
• Recordings and videos are deleted 6 months after they are created. Download and keep the videos you need.

Thumbnails

• Each time a version of the video is finished, Eizeno makes a thumbnail for it (a widescreen image like the video) with AI image generation. It appears below the video in "Finished video and revisions" on the canvas, and "Download thumbnail (PNG)" saves it. It is also used as the picture in the sidebar history (productions without a thumbnail keep the frame from the recording as before).
• The "…" button on each row of the sidebar history lets you rename that chat (leave the name empty to show the URL again) or archive it. Archiving only hides it from the sidebar list; the production, its video versions and completed videos are kept. Completed videos stay under "Completed videos", and you can open the chat from their detail page with "Edit in production chat".
• The thumbnail shows the product's name, a catchphrase taken from the explored site (its title, headings or description), the site's icon (logo) and the recorded screen, in colours that match the site. Figures, prices or ratings that aren't on the page are never added. Text from screens after signing in is not used.
• Ask in the chat, for example "change the thumbnail text" or "make it bolder", and the AI makes a new one. To choose the big text, write the exact words you want. Figures can only come from the explored page or your own message. Earlier thumbnails are kept and can be downloaded under "Earlier thumbnails" on the canvas. The video does not change.
• Because it uses the text and icon (logo) of the explored site, use it for your own service or one you have the right to use.
• If a thumbnail can't be made, the video is still finished as usual. No thumbnail is made when your balance is empty.
• On a share page, the version's thumbnail (if there is one) is shown before the video plays.

Share pages for prospects

You can turn a finished version into a page you send to prospects as a link. The page is a short service introduction that makes sense even without playing the video. From the top: the service name, its key point and a short description; the video (it never plays automatically); actual screens recorded during exploration (two or three, fewer if there aren't enough); and a button to the service (shown at the top and bottom with the same destination, with the destination's domain next to it). Viewers don't need to sign in. You can also skip the introduction layout, in which case the page keeps the earlier form: a title, the video, a short description and a button.

• In "Finished video and revisions" on the canvas, show the version you want and press "Mark this video as complete". This opens its details without publishing a link. To return later, choose the video under "Completed videos" in the sidebar. Versions you previously shared also appear here, so you can keep managing the same links.
• The top of the details page shows the video, its version and completion date, with "Save MP4", "Review share page" (when a share page exists; it opens a preview) and "Edit in production chat". To change the video itself, make a new version in the production chat; share page settings are handled separately.
• For a version without a share page, choose "Create a share page" and go through ① page content → ② button and publishing settings → ③ publish.

- ① Page content: a draft title, short description, service name, key point and actual screens (with headings and descriptions) is filled in. The draft uses only text from the explored site's public pages and the video's concept and plan; no track record, numbers or performance claims are added. Screens are taken automatically from the exploration recording (frames that passed the personal-information check; no tokens are used), and moments around a sign-in request and ranges the concept marked as unusable are not offered. Check and edit everything before publishing. With "Use the service's colours", the explored site's colours are used, adjusted so text stays readable. "Preview the page" shows the result right there before publishing. - ② Button: if you specified a button destination during production, that is used; otherwise the explored service's entry page and "Visit the service" are suggested. Parts of the URL such as the query (which can contain sign-in codes or personal information) are removed, so check that the destination is fine to publish. Also choose the page language (Japanese / English), how long it stays available and who can view. - ③ Press "Publish" to get the link. Pages are published only when you press the button; the AI never publishes one on its own.

• For a version with share pages, each share page appears as one card: the link ("Copy", "Open public page", "Preview"), the version, live / stopped / expired, expiry, who can view, the button destination, the page layout and a view summary. "Edit page content" lets you edit the title, description, button and introduction content in place and check them in a preview before saving (the live page doesn't change until you save). "Publish another share link" creates an additional link. Embedding, outcome tracking, version comparison and detailed reactions are in the secondary section under the card.
• Share pages published earlier keep their link and content. Screens taken later are never added to a live page unless you save them with "Edit page content".
• The shared version is fixed. When you revise in the chat and a new version is made, the video on a live page does not change. To share a new version, open it on its completed-video detail page with "Mark this video as complete", then press "Switch to version N" under "Switch a link from another version to this one".
• "Stop" makes the page and the video unavailable right away ("Resume" brings them back). The same happens when a page expires: neither the page nor the video is shown. You can extend the expiry, but only until the video is deleted (6 months after it is created).
• There are two choices for "Who can view". With "Anyone with the link", anyone who opens the link can watch. The link is hard to guess, but anyone it is forwarded to can open it, so it is not a way to check who someone is. For content meant only for specific people, choose "People who enter a passcode" and send the passcode separately. After repeated wrong attempts, entry is paused for a while. Changing the passcode means people who already opened the page must enter the new one.
• Only the person who published a share page can see or change its settings and engagement.
• The card's view summary shows, since the share page was published, page views, video plays, plays watched to the end and service button clicks. The share page and embeds are counted separately and never added together. When there are no views yet, or the counts could not be loaded, this is shown distinctly from 0. "Preview" and "Review share page" are not counted as views ("Open public page" is).
• "Engagement" in the secondary section shows how many times the page was opened, playback started, the video reached 25% / 50% / 75% / the end, and the button was clicked, by day, by source and by version. Add something like ?src=email to the end of the link to count those opens under that source name (lowercase letters, digits, - and _).
• These counts are aggregate estimates, not the number of people or deals. Repeat opens and reloads by the same person, and automated loads such as bots, can't be told apart and may be included. Viewers are not identified (no IP addresses or cookie IDs are stored).
• Direct lead form and CRM integrations are not available yet. Point the button to the booking or trial page you already use. To count sign-ups, bookings and purchases after the button, see "Track sign-ups, bookings and purchases" below.

Embed on your landing page or product page

You can embed a share page's video on your own landing page or product page. Open "Embed on your site" on the share page item, copy the "Embed code" and paste it into your page's HTML.

• The embed is the same as the share page. It uses the share page's version, expiry, stop, who can view and button, and "Stop" or expiry takes effect on your site right away.
• Until someone presses play, the embed shows only a still image and does not load the video itself. To keep your page fast, it loads only when it comes near the screen. It never autoplays, and sound plays only after the viewer presses play (the volume can be changed during playback). On phones it fits the width at 16:9.
• When the video ends, it shows the share page's button (such as a free trial or demo booking). While stopped or expired, it shows a notice and the button instead of the video (for share pages that need a passcode, the button is not shown either).
• Subtitles are part of the video image. Versions with subtitles show "Subtitles" before playback.
• If you list sites such as https://www.example.com under "Sites where it may be embedded", it is shown only on those sites (one per line; https://*.example.com covers subdomains). Leave it empty to allow any https site. This limits where it can be placed; it does not check who the viewer is. Use a passcode for content meant only for specific people — embeds ask for the passcode too.
• "Preview" shows how the embed looks. Opening and playing the preview is not counted in engagement.
• In "Engagement", embeds are counted under the source "embed", and you can also see counts by embedding site (only the site name such as https://www.example.com; page URLs are not recorded).

Track sign-ups, bookings and purchases

You can see how many people who clicked the button on a share page or embed went on to sign up for a free trial, book a meeting, buy, and so on, on your site. Your server sends these outcomes to Eizeno. Start from "Outcome tracking (sign-ups, bookings, purchases)" on the share page item.

1. Click "Issue a test key" (use "Issue a live key" for production). The key is shown only once. Save it in your server's settings and never put it in browser JavaScript. Reissuing a key stops the previous one immediately.
2. Once a key exists, eizeno_ref=... (a reference ID) is added to the button's link. The reference ID is a random value created on every click; it contains no name, email address or IP address of the viewer. On your site, keep this value, for example in a hidden field of your sign-up form.
3. When a sign-up, booking or purchase completes, send a POST from your server to the endpoint shown on screen, with the key in the X-Eizeno-Key header and a JSON body with these fields.

`` curl -X POST https://eizeno.com/api/share/<share ID>/outcomes \ -H "Content-Type: application/json" \ -H "X-Eizeno-Key: $EIZENO_OUTCOME_KEY" \ -d '{"type":"purchase","event_id":"order-10023","ref":"<the eizeno_ref you received>","amount":"19.80","currency":"USD","test":true}' ``

• type: signup / booking / purchase / custom (other)
• name: optional, for custom only: a name for the goal (lowercase letters, digits, _ and -, up to 32 characters; for example activation = the free-trial goal reached). Counted per name, apart from purchases
• event_id: your system's ID for the outcome. If the same ID arrives twice it is counted once (retries never double count)
• ref: the reference ID you received (omit it if you don't have one; the outcome is then counted separately as not linked to the video)
• occurred_at: optional. When the outcome happened, with a time zone (for example 2026-10-31T10:00:00+09:00). Defaults to when it was received
• amount and currency: optional. An amount ("12800", "19.80"; negative amounts and more decimals than the currency allows are rejected) and an ISO 4217 currency (JPY, USD, ...)
• test: true for tests. Anything sent with the test key is always a test. Tests are never included in live totals
• There are no fields for names, email addresses and the like (they are rejected)

How outcomes are counted (attribution rules)

• Only outcomes that happen within the attribution window after a button click (7 days by default; 1 to 30 days) and arrive with its reference ID count as direct outcomes.
• Outcomes without a reference ID, outside the window or with an unknown reference ID are counted separately as not linked to the video and are not credited to it. Outcomes from people who only watched and never clicked can't be counted.
• Changing the window doesn't affect clicks that already happened; they keep the window in effect at the time of the click.
• Under "Responses", each version and channel shows opened → played → button → direct outcomes, with the share of button clicks that led to a direct outcome, and the period covered (days the outcomes were received, Japan time).
• Amounts are the "reported purchase amount": the values your server sent, summed per currency. Refunds and cancellations are not subtracted, so it is not net revenue. Different currencies are never added together.

What can't be measured, and notes

• Outcomes can't be counted when, for example, the visitor leaves before your site stores the reference ID, signs up on another device or browser, or your site can't keep the ID because of consent settings or browser restrictions.
• The numbers are estimates and don't show that the video caused a sign-up or purchase.
• How your site handles the reference ID (whether it is linked to customer data, and how consent is obtained) is up to your site's privacy policy.
• Individual outcome records are deleted after 45 days, and reference IDs the day after their window ends. Daily totals are deleted with the share page's other totals when the video is deleted. "Delete outcome data" removes them at any time (view totals such as plays are kept). "Stop tracking" revokes the keys and stops adding reference IDs to the button.

Compare versions with different angles

From the same recording, you can make versions that open differently, show one of them to each person who opens the same share page (or embed), and see which presentation leads to more button clicks or sign-ups. Start from "Compare versions" on a share page.

1. Pick one hypothesis. For example: "Start with the problem, or start with the result?" Test only one difference per comparison.
2. Make a comparison video (if needed). In "Make a comparison video", choose the angle (start with the problem, start with the result, or a custom angle), confirm what will be made and that it uses production tokens, then press the button. Each press makes one video (nothing is made in bulk). Length, taste, features shown and the button stay as in the original; only the opening angle changes. The original video and its versions are kept. Like one edit, the tokens used are deducted at actual cost.
3. Start the comparison. Choose the hypothesis, the primary metric, the page opens needed before judging, and the versions to compare (2–3, same shape only), then press "Start the comparison". The primary metric is chosen before starting and doesn't change during the test.

- Button rate: views with a button click ÷ page opens - Measured outcome rate: direct outcomes (see "Track sign-ups, bookings and purchases" above) ÷ page opens. Available only on share pages with an outcome key

1. During the comparison, people who open the same link or embed see one of the selected versions. The same browser gets the same version where possible (the browser keeps one random value that decides the version; Eizeno doesn't store it and doesn't identify viewers; another device, another browser or cleared storage can get a different version). The title, description and button stay as set on the share page; only the video changes.
2. See the results. For each version: opens, plays, button clicks and outcomes, with the button rate and outcome rate. The period and the delivery conditions at the start (who can view, button destination, outcome tracking) are shown too.

How the judgement works (no winner from small numbers)

• Until every version reaches the number of opens you chose, it shows "Not enough data" and doesn't say which is better.
• After that, while the confidence interval (95%; wider when three versions are compared) of the difference includes 0, it shows "Can't tell yet".
• Only when it no longer includes 0 does it show "A difference is visible (for this period and these conditions)". This is not a promise that it will hold or that results will improve.
• Calculation: for each version, the rate uses the metric count plus 1 over the opens plus 2 (an adjustment so that the interval isn't too narrow with small numbers), and the difference and its interval are shown.

Swapping versions and ending

• If you swap a version during the comparison, it is recorded in the history, and the numbers before the swap are shown separately and not used for judging.
• Pausing or resuming the share page, or changing the button destination, who can view, allowed sites or the attribution window during the comparison is recorded in the history. Changed conditions make the comparison harder to read, so change them after ending the comparison where possible.
• "End the comparison" returns the share page to its original version (the version chosen on the share page). Results and history are kept and remain under "Earlier comparisons" when you start the next one.

Not enough or missing data

• When there are no views yet, the outcome keys are stopped, or events were not counted because of the daily limit, the panel says so.
• Outcomes can arrive until the attribution window ends, so the outcome rate reads low early in a test.
• Counts are estimates, not numbers of people or deals (reloads by the same person or bots can be counted).

Ideas for next time

• "Ask AI for next ideas" sends only the totals above to the AI and asks for presentation ideas to try next. Ideas with numbers that aren't in the totals are not shown. It uses a small amount of production tokens at actual cost.
• Eizeno doesn't change prices or place ads.

Services that need a sign-in

Send the URL and what you want to show in the chat. The installed extension connects automatically. The AI opens a dedicated new tab and pauses exploration and recording to ask you if sign-in is needed.

• After consenting in the chat (you can also just reply "yes"), sign in yourself in the dedicated tab Eizeno opened in Chrome. Navigation is limited to supported identity providers; unsupported configurations such as separate windows stop safely. When you are done, send "I signed in" in the chat (the same as "Finish sign-in and resume" in the extension popup). AI control resumes after Eizeno checks that the tab is back on the product and protections are active. If the tab is still on a sign-in page or an identity provider (such as Google), Eizeno keeps waiting: return to the product and send "I signed in" again. Never paste passwords or sign-in links into the chat.
• To stop signing in halfway, send for example "continue without signing in" or "only explore the parts that need no sign-in". The dedicated tab returns to a public page and exploration continues without sign-in. If that page turns out to be signed in, Eizeno asks before recording it.
• With an extension version before 1.8.0, "I signed in" in the chat cannot resume the same tab, so Eizeno explores again in a new dedicated tab that uses your Chrome sign-in (the earlier recording is kept), and "continue without signing in" stops the exploration there. Update the extension to continue in the same tab.
• The extension does not export cookies or saved passwords to Eizeno. After you start, the paired tab’s screen images and page text are sent for video production.
• Use a demo account without real customer data and with minimal permissions. The AI only views; it does not purchase, delete, send, save settings or sign in. Text input is limited to search and filters.
• Exploration and recording continue if personal information appears on screen. Some information is masked automatically, but Eizeno cannot guarantee removal of all personal information from the recording or finished video. Review the video before publishing or sharing it.
• If you decline before sign-in begins, exploration continues on public pages. If you do not finish manual sign-in in time, exploration stops safely; once you have signed in, send "I signed in" in the chat to explore again in a new dedicated tab that uses your Chrome sign-in (Eizeno asks before recording signed-in screens). Unsupported configurations or failed protection checks stop exploration. If sign-in is repeatedly required, request another exploration in chat.
• Stopping leaves your usual Chrome sign-in state in place. Sign out on the product if needed. Request another exploration in chat to start in a new dedicated tab.

Using Eizeno from Claude Code or Codex

From the Claude Code or Codex you use for development, you can tell Eizeno your product's public URL and what you want to highlight, and get an introduction video (this uses MCP, a way to connect AI clients to outside services). It is available when "Connect AI clients" appears on the Settings and top-up page.

How it works

• Eizeno's AI makes the exploration, taste and editing decisions. It follows the same flow as the web studio and uses production tokens the same way. The exploration runs in a dedicated tab of your own Chrome (the Chrome extension), just as on the web.
• The productions and versions also appear in the web studio, and you can keep revising them there.

Connecting

1. Under "Connect AI clients" on the Settings and top-up page, choose a name, a permission and an expiry, and press "Issue token". The permission is "View only" (status, versions, downloads, balance) or "Production too" (including explorations, edits and other actions that use production tokens).
2. Copy the token shown. It is shown only this once. Do not share it or put it in a repository.
3. Claude Code: run claude mcp add --transport http eizeno https://eizeno.com/api/mcp --header "Authorization: Bearer YOUR_TOKEN" in a terminal. In a team-shared .mcp.json, do not write the token itself; pass it through an environment variable such as ${EIZENO_TOKEN}.
4. Codex: add [mcp_servers.eizeno] to ~/.codex/config.toml with url = "https://eizeno.com/api/mcp" and bearer_token_env_var = "EIZENO_TOKEN", and put the token in the EIZENO_TOKEN environment variable.
5. Press "Revoke" for tokens you no longer use. Revoked and expired tokens stop working.

The settings format may change between client versions. Please also check each client's own documentation.

Example conversations

• "Use Eizeno to make a short introduction video of the new reporting feature for accounting staff. The public URL is https://(your product)" → your AI shows what will happen and that production tokens will be used; once you approve, the exploration starts → "Show me the tastes" → "Use the second taste" → after it is done, "Start with the problem" → "Give me the MP4 link".

Rules and limits

• Before any action that uses production tokens, your AI shows what will happen and that tokens will be used, and asks for your approval. Retrying the same request never produces or charges twice.
• Send only the product description and the points you wrote yourself. Do not send repository code, secret keys or passwords (Eizeno refuses text that looks like a key or password).
• Even for services that need a sign-in, Eizeno never takes passwords. You agree on the web studio page your AI client points you to, then sign in yourself in a dedicated tab of your own Chrome.
• Do not put numbers, prices or performance claims in the video that the product's screen does not show. Eizeno asks you to check such wording when it finds it.

Reviewing and revising in Slack

You can review finished videos, ask for revisions, and approve versions in a Slack thread. This is available when "Review and revise in Slack" appears on the Settings and top-up screen.

Connecting

1. Under "Review and revise in Slack" on the Settings and top-up screen, press "Add to Slack".
2. On the Slack screen, choose the workspace and the channel for notifications, then allow.
3. To change the channel later, press "Load channels" and pick one. Only channels the Eizeno app has joined can be chosen (type /invite @Eizeno in the channel).
4. When you no longer need it, press "Disconnect". Removing the app in Slack also disconnects it.

Notifications

• We post only when a video is ready, rendering fails, a job stops (for example, not enough balance), a taste needs to be chosen, or a sign-in is needed, in one thread per production. We do not post progress updates.
• The first message of the thread summarizes the versions, approvals, and revision requests, and is updated each time.
• Videos are never posted to the channel. The "Open in Eizeno" preview opens only for you, signed in to Eizeno.
• Only when you choose "share page" in Settings and have issued a share page for that production, the notification also includes the share page link (with that share page's expiry and passcode; it serves the version fixed when it was issued).

Asking for revisions and approving

• Only you, after linking your Slack account to your Eizeno account, can ask for revisions and approve. Being in the channel alone does not let anyone view or change productions or use production tokens.
• Linking: when you mention @Eizeno in a production thread without a linked account, you receive a link only you can see. Open it, sign in to Eizeno, check the workspace and the Slack user, and press "Link". Do not link unless you received the link yourself. You can unlink on the Settings screen.
• Revising: write something like "@Eizeno open by showing how easy setup is" in the thread. A confirmation only you can see appears; editing starts when you press "Revise" (it uses production tokens at actual cost; nothing starts until you press it). When the new version is ready, we post in the same thread. The same request never runs twice.
• Approving: press "Approve version N" in the notification to record that you approved that version. Approvals are recorded per version; when a revision makes a new version, the earlier approval does not carry over. Only the latest version can be approved.
• Eizeno never posts from Slack to other services (such as social media). Sharing a balance across a team is not supported.

What Eizeno can't do

• Sign in or create accounts by itself. Screens that need a sign-in are explored only after you sign in yourself (Services that need a sign-in). Passwords written in the chat are never used.
• Buy, pay, delete, post, or send. It does not move money, change data, or send anything out. It does not change settings either.
• With a compatible Chrome extension, Eizeno can follow a visible link to explore and record a manual or example relevant to the video, including one hosted on a different domain. Links that usually open a new tab open in the same dedicated tab. On other domains, Eizeno can browse public pages. Screens and interactions that require sign-in or JavaScript are not supported there. Eizeno skips links it cannot verify. Internal network addresses are outside the exploration scope.
• Fetch videos embedded from external services (such as YouTube), from other sites, streaming video, or DRM-protected video (How scenes are shown).
• Guarantee accuracy. The AI's descriptions may contain mistakes. Please review the video before publishing it.
• Depending on how a site is built, unexpected actions may happen or features may not be found. Anything it could not show is listed in the chat at the end of the exploration.

Costs and usage examples

Usage depends on the site, the number of steps, and the video. Measured examples (in production tokens):

• Exploring and recording a public web app in 17–25 steps: about 61,000–95,000
• One edit of the introduction video: about 40,000–105,000
• Adding an AI-generated opening shot: about 80,000 more
• Free-form videos (a free-form option, or your own description without an option): more, for the AI deciding the staging, building and checking the layout (depends on how many fixes the checks require)
• Thumbnails (Thumbnails): more, for one image per version and one more each time you ask for a new one (no measured example yet)

What you use is deducted at the actual cost of the AI services (in US dollars) × 162,162 production tokens. The studio chat assistant (The chat assistant) is deducted the same way (a very small amount per message). The support chat does not use production tokens.

Data retention

If an older production still contains extracted meeting notes or how-to guide data, you can delete it under “Previous production data” on the canvas. Existing videos are kept.

• Recordings, screenshots, introduction videos and their assets, and the records of explorations and edits are deleted after 6 months from creation.
• Your account and balance are kept until you ask to close your account.
• See the Privacy Policy for details.

Troubleshooting

"Your balance is too low"

There aren't enough production tokens to start exploring or editing. Top up from "Settings & top-up" and send again. Tokens expire 6 months after purchase, and expired tokens can't be used.

"An exploration is running"

Explorations, taste proposals and edits run one at a time. Wait for the running one to finish, or stop it with the stop button, then send again. An exploration that stays stuck without progress for a long time is closed as "failed" automatically after a while.

"This URL can't be opened"

Send the address of a public site that starts with https://. localhost, IP addresses, and internal addresses can't be used. For a service that needs a sign-in, send its home page or similar URL; the exploration asks you to sign in when needed.

An exploration stopped or failed

• "Exploration stopped. The recording so far is saved.": you pressed the stop button.
• "The exploration stopped because the balance ran out": check your balance, top up if needed, and send the URL again.
• "The exploration stopped because it reached the time limit (1 hour)": one exploration can last up to 1 hour. You can make a video from the recording so far.
• "The exploration stopped for safety because Eizeno could no longer confirm that the dedicated tab's screen was protected…": the page reached a state where Eizeno could not confirm that personal data is being hidden. If the recording so far is kept, choose "Make it from the recording so far" to make a video from it. If it is not enough, ask in the chat to explore again (say which pages to explore or avoid if you want it to skip the page where it stopped).
• "The exploration stopped partway": the site may have failed to load, or there may have been a problem on our side. Wait a little and send it again. If it keeps happening, contact us with the exploration's URL (the /runs/… address of the page) and the time. If a run stopped because of a problem on our side, we will check it and return the production tokens to your balance.

Editing the video stopped or failed

If you see "Editing stopped partway", choose a taste again or send your request again. If it keeps happening, contact us. If editing stopped because your balance ran out, top up and send again.

I topped up but my balance didn't increase

The balance is added when the payment service confirms the payment, usually within seconds. Reopen "Settings & top-up" to check. Some payment methods take longer to confirm. If it still isn't reflected after a while, contact us with the date, time and amount of the payment.

The checkout page doesn't open

If you see "The payment page could not be opened", wait a little and try again.

The sign-in link doesn't work

If the link can't be used, it has expired (it is valid for 15 minutes) or has already been used. Request a new sign-in link. If signing in with Google fails, try again or sign in with an email link.

"The job could not start"

We couldn't prepare the exploration or video rendering on our side. Wait a little and send it again.

Contact

• In the Contact chat, you can ask how to use Eizeno or get help with a problem. An AI answers from this manual; problems it can't resolve and feature requests are summarized and sent to our team (you confirm the content before it is sent). Signing in is required to use the chat.
• For problems, including the exploration's URL (/runs/…), the message shown, and the time it happened helps us look into it faster.
• Operated by ZenLab Inc. (Commercial Disclosure, Terms of Service)