You can review the video's plan before it is made — for 1 credit. Both the Visual step's 🎙️ Narrátorvideó and 🎬 Videógenerálás modes gained a Kreatív eszközök (creative tools) row with two tiles: Forgatókönyv (script) and Videó. Only Videó is on by default, so nothing changes if you were already using it. Switch Forgatókönyv on and Videó off, and within seconds you get a readable, scene-by-scene plan for 1 credit — read it, rewrite whatever you don't like, and the video is built from exactly that. The plan is edited on the left, above the finished video; the box opens full-screen and collapses too (so does the Vizuál box). Leave both tools on and the video costs the same as before. Your own text is never overwritten: with a filled field the Generálás button tells you why it has nothing to do, and offers a Mező ürítése (clear field) button. In video generation the plan is written in the language you picked under Szöveg és beszéd nyelve and needs Prompt optimalizálás switched on; for narrator videos it is written in Hungarian, and you also get the plan it used in the field afterwards.Help:Videó és narrátorvideó generálása
A new video engine to choose from: WAN 3 — and you can now make half-minute videos. The Visual step's 🎬 Video generation mode has a new Model menu. The familiar Grok Imagine stays the default (up to 10 seconds, all seven aspect ratios), and next to it sits the new WAN 3: 2–30 seconds, long enough for a real, watchable product showcase rather than a quick flash. With WAN 3 you can turn the sound on or off (Grok always adds sound and it cannot be switched off), but it does not know the 2:3 and 3:2 aspect ratios; if one of those was set, it switches to 9:16 by itself and tells you why. The price is unchanged: 1 credit per second, per video, on both engines. If a run would reach 30 credits, we ask first — so a long video never starts by accident. You can close the post meanwhile: the video finishes in the background and appears in the gallery on its own.Help:Videó és narrátorvideó generálása
The video's text and sound settings are simpler. In the Video block, the old Headline and CTA button switch pair (and their two text fields) have been replaced by a single switch: Szöveg megjelenítés engedélyezése (allow on-screen text). If you want specific wording on the video, you now write it in the Instruction — one place to say what you want, instead of choosing between two fields. The sound switch is now called Hang engedélyezése (allow sound). Your earlier settings are kept: if you had Headline or CTA switched on, text stays allowed.Help:Videó és narrátorvideó generálása
You now decide the order your blog articles are written in. Every row in the Kulcsszókutatás list gained a drag handle: grab it and move the keyword up or down (touch works too), or use the Legfelülre button at the end of the row to send a keyword to the top in one click. The order saves itself and survives a reload. To make the machine follow it, the keyword set's knowledge source gained a new mode: Sorban (felülről lefelé) — every article is then written from the topmost approved keyword that hasn't been used yet, and the used one drops out on its own. New sets already default to this; older ones need switching over once. Until you drag something, the list keeps its usual order, and keywords arriving from a later research round go to the end — they never disturb the priorities you set.Help:A cikkek sorrendjének beállítása
You can now see which of your Facebook posts drove traffic to your shop. Until now the Top posztok list in Analitika contained no Facebook posts at all — they now show up, with the real number in a new Átkattintás column: how many people clicked on the post. For a webshop that says more than likes do. Megtekintés and Interakció stay empty on Facebook for the moment ("—", not zero): those need a separate permission from Facebook, which we're working on — the ? next to each column name explains which number is missing and why. Nothing changes on Instagram. The list reserves room for both platforms, so your Facebook posts stay visible even when your Instagram numbers are higher. The first numbers arrive with the next daily refresh.Help:Analitika: hogyan teljesítenek a posztok·Mit jelentenek az Analitika számai
A new image model is available: GPT Image 2.5 Sunburst. A fourth row appeared in the Modell menu of the Vizuál step. Pick it when you want exactly one thing changed on an existing image while everything else stays put — it sticks closest to what you ask for. It costs the same as GPT Image 2.5 Flare (5 credits), it is just slower. It's available in the post's Vizuál step (both Létrehozás and Szerkesztés), in bulk editing, for newsletter/blog body images, and in Ninja AI.Help:Kép generálása a poszthoz
Changed
New posts now start with the Új kép tool switched on. Until now no creative tool was selected in the Visual step, so the first Generálás only popped up a window asking you to pick one first. From now on 💡 Új kép (new image) is on by default — switch it off in one click if the post needs no image. Your existing posts, templates and automations keep their settings.Help:Kép generálása a poszthoz
GPT Image now produces sharper images, far faster — for the same credits. The old _GPT Image 2_ entry in the Modell menu is now GPT Image 2.5 Flare: same place, same 5 credits, better image and about an eighth of the wait (measured: 20 seconds instead of 150). There is nothing to switch over — your saved posts and templates use the new model on their own.Help:Kép generálása a poszthoz
Image editing now runs at the highest quality — for the same credits. The Szerkesztés mode in the Vizuál step (and body-image edits) used to run one quality step lower to save cost; with the new model that's no longer necessary, so we switched to the detailed setting. Same 5 credits, nothing to change on your side — edited images simply come out better. It takes a few seconds longer in exchange.Help:Kép generálása a poszthoz
Fixed
Ninja AI can now see and analyse your previous ad copy — in any assistant. Until now, asking it to "analyse my previous ads" got you a reply saying it had no access: only the General assistant could reach your ads, the Prompt writer and the Analyst could not, and none of them returned the Leírás field. Now every assistant lists your ads with all three text fields (Főcím · Elsődleges szöveg · Leírás), in batches — and tells you how many it worked from. The Prompt writer turns them into a general, variable-driven ad prompt (without burning in one specific product), and the General assistant tries it out right away on an ad, so you can see in one round whether it beats the built-in one. An ad link pasted into the chat (`/ads/<code>`) is recognised too.Help:Hirdetés-prompt írása: a három mező
A blog template built from the chat now really does what Ninja AI says it does. The assistant sometimes claimed it had turned on Anti-AI mód or set the research mode, while the template was in fact created with factory defaults — because its tools had no field for those settings at all. The template is now created in one step, with its full configuration (Anti-AI, research mode, topic instruction, search language, knowledge source), and the approval card shows those in plain language before you confirm. On top of that, post-shaped settings can no longer be written onto a blog template: the system tells you which tool is the right one.
One keyword research no longer produces two knowledge sources. The knowledge source is created automatically when a research round closes, but Ninja AI used to add a second one alongside it — two identically named rows in the picker, with no way to tell which was real. It now uses the existing one and names it in its reply. Renaming the research now always carries the name over to the knowledge source (with long names and manually renamed sources it never did), and deleting the research takes its knowledge source with it. If an old source still points at a deleted research, we now say exactly that — instead of telling you to "approve some keywords".
