Studio
Where a video is made on a canvas of steps — a presenter, the line, the prompts, a key frame and the clip — each one you can run on its own.
The Studio has three tabs: All, Videos and Photos. New workflow starts a video, a picture or one from a template, and each opens on the canvas described on this page. Photos is your image library. The Video Producer makes videos the same way when a batch comes round or when you ask it in a chat; this page is for making one by hand.
The canvas
A video is a diagram. Cards on the left are what you bring: a presenter, the line they say, a product from your catalogue or a photo, an element such as your logo or a screenshot, an on-screen text (a hook, a caption or a news headline), and your brand kit. Cards to the right are the steps QuQi runs: a text step writes a prompt from what is wired into it, an image step makes a picture, and the video step makes the clip — the model speaks the line itself, so there is no voice-over to record and no lip-sync.
Wires say what a step reads. Drag from a card’s dot onto another card to wire it in, or press the + on a card and pick from the list. A wired text step flows into an image or video prompt as {its-name}; a picture wired into the video becomes its first frame or a reference for the person and the product. Every card is named in lowercase with dashes, and that name is exactly what other steps read.
- Right-click the empty canvas to add a node where you clicked. Drag cards anywhere; scroll to zoom; hold shift to pan; Arrange lays everything out left to right from the wires.
- Double-click a name to rename a node. Every wire and placeholder that pointed at it follows.
- Undo and redo cover every edit, twenty steps deep.
Templates and the Copilot
A new video starts from a template: Talking head, Talking head with hook, Product in hand, Conversation (two presenters, two lines), Podcast (two people at a studio table with microphones) or News desk (two presenters at a desk, with a lower-third headline). The two-person templates hold the whole conversation on one card, a turn per line starting with A: or B:, written by the Copilot or Write it with as many turns as fit; they make one picture of both people together in one room, in the clip’s shape, and one take shot like a two-camera recording: no camera movement, the picture cuts to the camera on whoever is speaking, two voices kept apart. Each line card says its part while it is empty, and Write it writes it in that part. Every picture step has a shape of its own in its panel; the clip’s shape is set on the video step. Save any canvas you like as your own template and it appears beside them.
A line card has Write it: one call writes what the presenter says from your website’s own context — the business, the brand voice, who it is for — and, on an exchange card, the whole back-and-forth. One credit, and every word stays yours to change.
The Copilot fills a canvas from one sentence — who is on camera, what they say, what it is for. It writes the line, the prompt steps and the hook, and picks the cheapest model and length that fit. Read it through, change what you like, then run. A fill is one credit.
Running it
Models and length
The video step’s model is a dropdown on its card, each option priced for that step. Seedance 1.5 Pro is the standard; Kling v3 Pro, Seedance 2.0, Seedance 2.0 Fast, Veo 3.1 and Gemini Omni Flash are there when you want their look. Each model has its own lengths; the length snaps to what the model can make, up to 12 seconds, and the line is cut to what fits at speaking pace.
On-screen text
Add an On-screen text node, write the words, pick the style — a POV hook across the top, a caption along the bottom, or a news headline on a lower-third bar — and wire it into the video. The style is a starting point: the node’s panel sets the size, the padding, the background and text colours, the alignment and the exact place in pixels of the frame, and the clip card previews the result. By default it is burned over the whole clip at the finish, spelled exactly as you wrote it, and changing it later re-runs only the finish, never the clip — the button says Burn text when that is all that is left. You can instead have the video model draw it in the footage itself: the words become part of the clip, which looks more natural but is not promised to be spelled right, and a change re-renders the clip.
What a video costs
Videos are paid for with AI credits — the same monthly pool the image library draws on.
- One credit is one AI call: a prompt step or a picture.
- The clip costs its seconds times the model’s rate — one credit a second on Seedance 1.5 Pro, three on Kling v3 Pro, seven on Veo 3.1.
- A 12-second talking head on the standard model is 15 credits: the clip, the key frame and two prompt steps.
- The Run button shows the total before anything renders, and a step that already rendered is never charged again.
If a video costs more than you hold, the canvas says so before anything starts: how many credits it needs, how many you have, and that you can buy a credit pack or pick a shorter clip or a cheaper model instead.
When it is finished
A finished video stays in the Studio, where you can download it. To post it, ask in a chat — “post this video to Instagram and TikTok” — and the Social Copywriter turns it into a post, asking which channels and whether it should write the caption. The post goes out, or waits for your OK, as its auto-approve says.