The video wizard explained: Cast, Shots, B-roll, Studio, Writing

One flow, three templates

Every video walks the same spine: Template, then who is on camera, then Source, Script, and a build step before the render. The template decides which build steps appear:

• UGC Reel: Template, Avatar, Source, Script, Shots.
• Two-Hander Podcast: Template, Cast, Source, Script, Studio.
• Split-Window Reel: Template, Avatar, Source, Script, B-roll, Shots.

The stepper at the top always shows the steps of the template you picked. UGC Reel is available on Starter; the podcast and split-window templates show a Creator+ pill and open an upgrade prompt on a lower plan.

If you want to see the finished product before you read the steps, that page shows what each template produces.

Template

The first screen, “What are we making today?”, opens with two doors side by side.

• Recreate a reel, on the left. Paste an Instagram or TikTok reel link and click Recreate →. A recreated reel walks its own steps, including What do they wear?, which dresses each avatar in one of their own outfits by default. Full Clone and Transcribe covers them.
• No link yet?, on the right. See what’s trending opens Viral Reels, where you can pick a reel that is pulling views in your niche and come back with it. Tracked Competitors and Viral Radar explains that page.

Below the doors, one card per template with its name, tagline, and “from ~N credits”. Clicking a card picks it and moves you on.

Avatar and Cast

UGC and Split-Window use the Avatar step: one avatar on camera. The podcast template uses Cast instead: tap one avatar for the host (seated right), tap another for the guest (seated left). Tapping the host again clears the pick. If you own only one avatar, the “AvatarFactory co-hosts” rail seats a house co-host as the guest. Continue to Source → unlocks once both seats are filled.

Source and Writing

Source is where the script comes from: the AI content expert, a pasted reel, a followed account, or a blank page. When you pick an AI hook, the wizard adds a short Writing step, “Writing your script.”, where the writer expands the hook. It moves to the editor by itself when the script lands; ← Pick a different idea takes you back.

Script

The editor, the word count, the seconds estimate, the Hear It preview, ↺ Revert, and the Script Co-writer on the right. In the podcast template, speaker tags name who says each line. Template flows can list script rules under the editor; blocking rules disable the continue button until they are fixed, and the co-writer offers to fix them for you.

Shots (UGC Reel)

The Shots step, “Build your shot puzzle.”, is where the video takes shape. The AI producer reads the script and directs a first cut; the banner above the list shows its progress, then offers Re-direct All for another pass and an undo when you want your last cut back.

Each shot card holds:

• The model. The cinematic model (“Cinematic motion, camera moves”) or the talking-head model (“Talking head, lip-synced”). Every plan can pick either one. Render times, models and the receipt explains both.
• The scene. A saved scene from your avatar, a new one generated in the scene drawer, or “Starts where the previous shot ends”, which chains the shot to the last frame of the one before it. When a shot has no scene yet, the Producer Suggests panel describes one. On the first shot it offers Generate This Scene. On every shot after it, three choices: Generate This Scene for a new one, Continue From Previous Scene to keep the same set, outfit and framing as the shot before, and Pick an Existing Scene for your saved scenes. Continue stays greyed out with “Generate the previous scene first” until the shot before it has a scene. When that last frame shows no face (the shot ended on your hands, or from behind), the next shot starts from the scene the chain began on instead, so the render always has your avatar’s face to work from.
• The direction and the transcript. What the avatar does and the words for this slice. Audio tags in brackets shape the delivery and are never spoken or captioned.
• The length. A talking-head shot runs 3 to 120 seconds; a cinematic shot runs 3 to 10 seconds.

Above the list, the length meter reads “Video length your plan renders” and turns into “Over the length your plan renders” when the cut is too long. Shorten to Fit trims it and offers Undo for a few seconds. Subtitles is a single On / Off switch for the whole video; turning it on opens “Choose a caption style” with Studio Pop, Chip, Stamp Drop, Marker Swipe, Mono Bracket and Typewriter.

If some shots still need scenes, a panel offers Generate N scenes or Generate scenes + video, which generates the scenes and renders in one go. The footer button reads Generate · N cr and asks “Start the render?” before charging.

On a trial, every step above is yours to use and the trial renders one video. After that video, the plan is asked for at that button: the line beside it reads “Your trial includes one video. Upgrade to keep generating.” and both Generate and Generate scenes + video open a card titled “Render Your First Video” over the step, with Unlock Rendering With Starter (or whichever plan you are on) and a Keep Editing link that closes it and leaves the shot list untouched. Paying there brings you back to the cut you already built with the “Start the render?” confirmation open, so one click starts the render.

B-roll and the talking window (Split-Window Reel)

The split-window template cuts the script into two to five chapters. The B-roll step shows one clip per chapter that we matched to your words, plus a music bed. On each card, Change searches the licensed library and Generate edits the AI clip description; ↺ Our pick restores the suggestion. Matching takes about a minute.

The Shots step for this template, “Frame the talking window.”, has one decision: the image behind your avatar. Choose Natural backdrop, one of your saved scenes, or Generate a scene. Then click Animate → to render.

Studio (Two-Hander Podcast)

The Pick your studio. screen lists your studios, presets and past episodes. Generate New Studio opens the studio agent: describe the room in your own words, and it writes the brief, asks you a question when it needs one, and returns empty studio options (the cast is seated later). Click an option to change it in plain words, then Use this studio. Below, add an optional prop behind each person and click ✦ Seat the cast · ~6 scene images to see both people in the room. Each seat has ↻ Re-roll. Animate → renders the episode once the cast is seated.

Render and Shipped

The oven screen confirms the render is running on our side. When it finishes, the You Shipped screen shows the player, the render receipt and Download. Render times, models and the receipt reads the receipt line by line.

On a trial, the upgrade prompt appears here, on the Generate click, and nowhere earlier. You build the whole video first: template, avatar, source, script and shots. Nothing is charged before you pick a plan. Create your account and choose a plan covers the plans.

Drafts and the step guard

The wizard saves your draft on every step change and after each pause in typing. A reel you recreate is saved the moment you click Recreate →, before the analysis finishes, so leaving the page during it costs you nothing. Unfinished drafts live under Drafts on the Videos page, and a plain visit to the wizard offers to resume the newest one: “You have an unsaved video draft from 5 min ago”, with Resume draft and Start fresh.