Gobinar opens your application on an isolated machine, clicks its way through it and builds a demo video with voice-over and subtitles. Every sentence is tied to what was visible on screen while it did. No filming a monitor, no editing suite, no agency.
Example recording. This is what it looks like when Gobinar works.
A proper product video from an agency costs a few thousand euros depending on scope and takes weeks. The real problem is not the price, though — it is the shelf life: by the release after next it shows an interface that no longer exists. So it does not get replaced; it quietly stays in use, or never gets made at all.
People fill the gap instead. Someone from sales or technical consulting clicks through the same seven screens yet again because a prospect wants to „just see it once“. That is the most expensive way to deliver a demo, and the only one that does not scale.
Gobinar turns that around: the demo comes out of the software itself. If the interface changes, you have it regenerated — instead of reshooting it.
The steps marked in colour are approvals. They cannot be skipped and cannot be switched off — at Gobinar there is no video about someone else's software that comes into being without a human yes.
The address of your software, who is watching, what the demo should achieve and how long it should run.
A free pre-check without a browser: reachable, encrypted, no internal target. If it fails, nothing is charged.
An isolated machine opens your software. A model looks at screenshots and proposes the next step — clicks go through labels, never through pixel coordinates.
You see every scene and every step that touches anything — and also what was discarded. Nothing proceeds without your yes.
The approved plan is played back. Without a model, so it is identical every time.
You see the recording before it is narrated.
Two to four sentences per scene, narrated and cut together — with subtitles from the same timeline.
The last click is yours. There is no automatic „it is done, off it goes“.
A video about your software that shows something your software cannot do is worse than no video. Three mechanisms make sure that does not happen — none of them is a statement of intent.
Every step remembers what was clicked — test identifiers, role and name, label, visible text. Coordinates appear nowhere. That is why the recording still works when the layout shifts, and why nothing ever clicks into empty space.
The model may explore and propose. What is then recorded is the approved plan — played back like a script, without an AI deciding anything again. Two recordings of the same plan produce the same result.
The narration is generated per scene from what was visible in the screenshot. A check over it discards anything that goes beyond that: invented metrics, third-party systems, superlatives. Rather than guessing, the plain scene heading is left standing.
The target was a publicly reachable inventory-management system with seven linked screens on someone else's server. The model found its own tour: home, overview, orders, stock, customers, reports, delivery quality.
What was not good about it yet: Three of the ten scenes carried generic headings, because the model requested a screenshot without a label there. As evidence that the chain holds, that is enough — as a sales video it is not. We write that down instead of leaving it out.
We know which question comes next here, because we would ask it ourselves. So, plainly: these are measured properties, not assurances.
Every run gets its own machine, which is torn down automatically afterwards. Measured: four runs, four teardowns.
Only the exact image whose checksum is on file will start — no „latest version“.
No administrator, an immutable file system, all special privileges removed, no privilege escalation possible.
The machine has no network access of its own. Every connection goes through a forced proxy that lets through only the address you approved — encrypted, standard port, nothing internal.
Results come back over a dedicated channel, not over any access to Gobinar.
The recording machines run exclusively in Falkenstein and Nuremberg.
Recordings are not public. Every retrieval needs a freshly signed link that expires after 15 minutes — plus your permission on top.
No forced player, no embed code, no dependency on us. You get the video and the subtitles and do what you like with them — website, sales email, trade-show booth, onboarding.
No surprise on the invoice: before a run starts, the total is there — itemised.
The expensive part: the model looking at your software.
Playing back the approved plan, machine runtime.
Recording the narration, with a timeline for the subtitles.
Cutting it together into the finished video.
If a run fails, the reserved credit is returned in full. For the same recorded material you get up to three attempts at narration and editing.
Today Gobinar records what is reachable without a login: your public application, a demo environment, a test system with invented data. That covers by far the greater part of what a sales video is needed for.
The second mode — the browser signs in to a protected system and records there — is fully built and has been through several security reviews. The step where you hand us a logged-in session, however, is not released yet: a complete logged-in run is not possible today. The reason is not an open task but a limit of the method — one of the protective layers reads information from the recorded page and can therefore, in principle, be deceived. We will release it when we are satisfied with it, not when it would sell.
If your use case strictly requires a login, schreib uns. We will tell you honestly whether and when that will fit, rather than selling you a subscription.
The runs per month are calculated from the 570 credits, not estimated. What you do not use, you do not pay for twice — credit packs can be added at any time.
A demo from your running software — you record nothing yourself. To get started.
No Gobinar branding, embeddable, for a team — the standard.
White-label, more runs per month, Enable included.
Larger teams, specific requirements or contracts? Talk to us — or first take a look at all plans an.
No. Today Gobinar records whatever is reachable without a login — your public application, a demo environment, a test system with made-up data. We neither need nor read your source code. For the mode in which the browser signs in to a protected system, the decisive step — handing over an authenticated session — is deliberately not released yet. We will release it when we are comfortable with it, not when it would sell.
The text is generated per scene from what was actually visible in the screenshot. A check then runs over it and discards sentences that go beyond that — concrete numbers with units, third-party systems, superlatives. Whatever fails is replaced by the plain scene heading rather than guessed. And you read it over at the end anyway.
The reserved credit is returned in full. For the same recorded material you get up to three attempts at narration and editing.
15, 30 or 60 minutes. The longer ones do not cost proportionally more but scale by a factor of 1.8 and 3.2 respectively — because exploration and preparation happen only once.
As MP4 at 1920×1080 with 30 frames per second, H.264 video and AAC audio, prepared for streaming. Plus a WebVTT file with the subtitles from the same timeline.
Creating an account, creating a project, connecting your software and running the pre-check: yes, free. A real recording costs credit — 570 credits for 15 minutes. The free account comes with 150 credits, so for your first real run you need a plan or a credit pack.
Create an account, enter the address, run the pre-check — that costs nothing and takes minutes. Only the recording run spends credit, and you see exactly how much beforehand.
No credit card · servers in Europe · you approve every step