Every look in your film, now one click away

The looks we shipped this morning store five settings, and the editor showed only one of them. The rest could be changed by an AI assistant talking to our API, which is not where most people tweak a film.

So we set one rule: every option the renderer can draw is reachable the same way through every door. The editor, our REST API and the assistant tools take the same fields, and a finished film reports back the full style it was made with.

In the editor, the Style panel opens with the looks as five cards. Below them, each setting is a labelled control with one line saying what it does, in English, French, Spanish and German. Picking a look resets everything to that look's choices; changing one control changes only that one.

Alongside the panel, we filled the shelf people expect from a video editor:

• Colour grades: warm, cool, vivid, mono, vintage and contrast, plus none. A grade touches the footage only, never the type, so in mono the page drains to grey while your captions keep your brand colour.
• Fifteen transitions, seven of them new, from slides and a spin to a light leak, a film burn and a zoom blur.
• Caption animations: pop, bounce, wave, glow and slide, held within bounds so a moving word never leaves the safe area of the frame.
• Emphasis: words your script marks stand out, or, in auto mode, every figure, price and percentage.
• Callouts: a ring, an arrow, an underline or a dimmed surround that points at the subject, each with a dark edge so it never vanishes into a page of the same colour.
• Cursors: the system arrow, a pointing hand, a presenter's dot, or no pointer at all.

One detail matters more than it looks. Some transitions need the previous scene to stay visible underneath while the next one arrives. That rule is now worked out from each transition's own movement rather than from a hand-kept list, so every new effect behaves correctly from the first frame.

The last item is a closing card with a QR code: the shortest path from "I want that" to your page. A video frame needs a plain grid of dark and light squares, and the QR libraries we found were built for web pages, so we wrote a small encoder of our own with no dependencies.

A QR code can look perfect and still not scan, so we test ours with a real decoder rather than with our eyes. Five codes of different sizes are decoded, then decoded again with a logo over the middle, and codes are read back out of rendered frames in landscape and portrait. The tile is sized from the actual code, which keeps the quiet margin scanners need.

The closing card can also show your social handles, your logo alone, or nothing at all, and the watermark can sit in any corner at any strength.

Passing tests is not the same as a film looking right. Before calling the day done, we rendered a real film with every option named and read it frame by frame. Two small details caught our eye, and both are now polished.

The spoken word in a caption grows slightly so you can follow along. On a long word, that growth reached into the space beside it, and two words seemed to touch. The enlargement is now capped to the room the gap can give, which also suits the eye: a long word needs less help to stand out.

On a scene whose camera framed a whole hero section, a callout ring traced the edges of the entire picture. A mark exists to single something out, so a callout now steps aside when its subject already fills most of the frame.

The lesson of the evening became a test. It follows each of the nine style options from the look's table, through what is stored, through what the pipeline builds, to what the renderer reads, using the real code at every step. A second half proves that the editor's style code never pulls the video renderer into your browser, and we checked that it fails when it should.

The result for you is simple: what you choose in the panel is what appears in the film. See the options on the features page, or start a film and open the Style panel.