How to Use Moosky Project Director to Make Your First AI Video (Step-by-Step)

Most people's first AI video dies at the first screen. You open a tool, find a text box or a wall of model names, and realize you have to know how to prompt, how to plan shots, and how to keep twelve separate clips looking like one video. Project Director skips that cold start. It behaves like a producer you brief in plain language: it writes the script, lays out the scenes, shows you what the run will cost, and renders the finished video once you approve. This walkthrough covers the exact click path, what each screen does, and where credits come in, using a 30-second product teaser as the worked example.

What Project Director actually is

Project Director is Moosky's multi-scene video pipeline. Instead of generating one clip at a time and stitching them yourself, you hand it a brief. It produces a script broken into scenes, generates each scene's video with shared characters and settings so the shots belong to the same world, and assembles everything into a single MP4 you download from the project page. You stay in the loop at two checkpoints: reviewing the script before anything renders, and reviewing the takes after they do.

If you want the background on why multi-scene planning matters versus one-shot generation, our complete guide to AI video generation covers the workflow categories the Director sits on top of.

Step 1: Open the Director and brief it

Open the Project Director. You get a chat box and, below it, a row of playbook starters: Product ad, UGC ad, Story, Explainer, and Narration. Each starter preloads a structured brief you can edit, which is the fastest way past the blank-box problem. Every link above opens the Director with that starter already selected. You can fill in the brief before signing in; the Director asks you to sign in only when you send it, and it keeps what you typed.

For our example, a 30-second teaser for a ceramic-lined travel mug, the Product ad starter is the right pick. Its prompt is roughly "Create a 30-second product ad using my brand and product assets, with a bold, premium feel and a clear call to action." Edit it to add what the Director cannot guess:

• What the product is, in one concrete sentence, plus the feature that matters most (ceramic lining, no metallic taste).
• Where it appears: kitchen counter, commute, office desk. Locations are cheap to vary; invented brands are not.
• The CTA and any text you want on or off screen. Say "no on-screen text" if you plan to add titles later yourself.
• Any assets you have. You can upload product photos in the first message, and the Director conditions shots on them.

You do not need shot lists or camera language at this stage. The planner does that. You need facts about the product and the intent of the ad. The UGC and product-ad variants of this structure are covered in depth on the UGC ads and product ads playbook pages.

Step 2: The plan conversation, and the price tag

The Director replies with a proposed script and scene breakdown. Read it before doing anything else, because this is the cheapest point in the whole process to change your mind. You can ask for edits in plain chat: "shorter hook", "drop the office scene", "make the last scene a close-up". Revisions here cost planning credits, not production credits, so iterate now rather than after the render starts.

Alongside the plan you get a cost estimate for the production run. It prices the whole project at your selected output settings before you commit, which means budget surprises happen at review time, not after the credits have burned. If the estimate is higher than you want, trim scenes or drop the resolution now, in chat, rather than mid-generation.

Step 3: Script review, and the credit reality check

When you approve the plan, a Script review panel opens. This is the screen people skip, and it is the one that saves the most credits. What it shows:

• Scene list. Each scene is a collapsible row with a name and a duration. Expand it and you get the scene's description, its cuts with camera actions and timings, plus chips showing which characters and settings appear in it. Those chips are your continuity check: if scene 6 shows a character nobody cast, or a setting that does not exist yet, fix it here with a change request.
• Output selector. Resolution and orientation for the whole project. Portrait for Shorts and Reels, landscape for YouTube and embeds. Changing this later means re-rendering, so decide once, here.
• The credit line. Just above the approve button, the panel states the credits the run needs and your current balance, in the style of "Needs ~N credits to start production · You have X." This is deliberate: production has a real admission floor, a project needs at least 500 credits in your wallet before the Director starts generating scenes, and the panel tells you before you approve rather than refusing after. If your balance is under the line, the figure turns amber and an Add credits link appears next to it. The signup welcome bonus is there to let you plan and test; it is not sized to cover full productions, so check the number, not your gut.

You have two buttons: Request changes, which sends you back into the planning conversation, and Approve & start production. Approving saves the scenes to your timeline and kicks off generation. Approving and starting production at the estimated cost are the same click, so the credit line is genuinely your last checkpoint before spend.

For our travel mug example, the review catches two things worth fixing: a fifth scene nobody asked for, and a kitchen setting with the wrong light. One chat message fixes both, and the estimate drops when the scene goes.

Step 4: Scene layout and takes

After approval, the project page becomes your edit surface. Scenes appear on a timeline, generating as the pipeline works through them. Per scene you get:

• Preview modes to watch the cut assembled versus individual scenes.
• A takes strip. Every generation of a scene is kept as a selectable version, so a scene that mostly worked does not need a full re-render of the project. Pick the take you want.
• Reshoots in the thread. If a scene is wrong in a describable way ("the logo should face camera"), ask the Director in chat and it regenerates that scene with the fix, keeping the rest of the project intact.

Watch with an editor's eye, not a viewer's. The failures at this stage are specific: a product photo that drifted in color, text baked into a shot that you wanted to add later, a scene that runs four seconds when its beat is two. All of them are cheaper to fix one scene at a time than to re-render the whole thing.

Step 5: Render and download

When the takes are selected, the review canvas switches to Final render, and the project renders to a single MP4 you download from the project header. If you edited anything after a render, re-render before sending the file anywhere: the download reflects the last render, not your latest edits. For the mug teaser, one 1080p landscape render plus one portrait pass of the three strongest beats covers the places a first ad actually gets shown: the landing page, YouTube, and a vertical cut for Reels.

What to expect the first time

Two honest notes. First, timing: a multi-scene run is not instant. Planning is quick, script generation is quick, scene generation is where the minutes go, and the timeline shows progress per scene. Second, credits: the system is designed so you see costs before you commit at every step, plan estimate before production, credit requirement before approval. Budget for one revision cycle on your first project. The first video is rarely the keeper; the second one, briefed with what you learned from the review screen, usually is.

Make your first video in the Project Director — brief it like a producer, review the script before it renders, and walk out with a finished MP4.