Video belongs on the canvas
A film in Shyne is a sequence of frames with time, not a separate editor you jump to after the deck is done.
If video lives in another app, the brand, the copy, and the stills all have to be rebuilt. In Shyne, a film is a sequence: the same frames you already designed, with playback and time attached.
That is the entire argument. A deck and a film are not cousins who meet at export. They are the same kind of object with a different clock. An ordered sequence has rank. A timed sequence has rank and duration. The teammate that wrote the title card can also place the clip under it. You can pull one frame back into a spoken deck, or push a slide into a fifteen-second bumper, without changing tools.
We did not add a “video mode” that hides the canvas. We let a sequence carry timeline state. If you can already edit a frame, you can already edit a scene.
Frames first, time second
You still start with something you can see — a scene, a title card, a product shot. The sequence puts those frames in order and gives them duration. Motion, voice, and generated clips attach to that structure instead of replacing it.
A timed sequence holds video-wide playback settings, picture tracks, audio tracks, markers, and ordered frames. Clips can be trimmed, sped up or slowed down, faded, and mixed. Design motion — reveals, transitions, animated type — lives with the composition. Clip timing lives on the timeline. Those are different jobs. Mixing them into one mystery slider is how people lose an afternoon.
Plan sound with picture. Do not add it as an afterthought unless you like repairing an edit that was already locked. Several short, intentional cuts usually read better than one long take. That is editorial advice, not a product constraint. The product will let you make the long take. The teammate, if you ask it well, will try not to.
Source footage and sound from stock search or from project files. The file library is the same library the deck used. The logo does not need a second upload with a different color profile because the video tool “likes it that way.”
Build an edit you can still open
Lay out clips on the timeline. Keep the frames selectable. An edit that can only be scrubbed as a flattened movie is an export you made too early. Shyne’s prejudice is the opposite: keep the graph, compile the MP4 when someone needs the file.
Scrub in the editor. Ask an agent to inspect exact moments as stills. When a title is late, you are not guessing at a render. You are looking at a frame at a time. Agents can inspect the edit, place and trim clips, and check exact frames through the same MCP lane they use for stills. The working loop does not change: orient, read, change, look, report.
If the frame looks right in the editor and wrong in a screenshot or export, check the warnings the tools returned. Keep styles, scripts, and media local to the frame. External references the renderer cannot load are the usual cause. Video does not get a free pass on that rule. A font that lives on a laptop and a clip that lives on a forgotten URL will both fail the compile.
Motion is not a different document
Animated type on a title card is still a node. A transition between two scenes is still a relationship between two frames. We are building motion as something the composition owns, not as a side project that forgets the brand.
If you have used tools that treat motion as a one-way render, you know the tax. You change the word, you re-render the world. In Shyne the word is still a word. Change it. The motion that was attached to it should still have a node to hold.
This is also why video belongs next to the deck. The launch sequence already has the words. The bumper needs three of them, larger, for two seconds each. That is a new sequence, not a new company.
Export when it is ready
MP4 compiles from the current graph. Every plan can export video. Hosted renders use the organization’s credit pool. Generating new AI video is a paid-plan media path. Exporting the film you already built is not locked behind a watermark.
Those two sentences are easy to confuse, so we will keep them separate. Free can export MP4. Paid plans unlock additional AI image and video models used to generate new media. If you brought your own footage, you are in the export conversation, not the generation conversation.
Desktop can fulfill the same export contract locally that the cloud fulfills on the server. The artifact type does not change because you clicked from the app on your machine. PNG, PDF, HTML, PowerPoint, and MP4 are compile targets. They are not separate editors.
Why this is the demo
A generator that hands you a clip is common. A teammate that designs and films on the document you are looking at is not. That is the version of video Shyne is built to keep: editable, on brand, and sitting next to the rest of the work.
The demo we care about is not “watch a model invent a city.” It is “watch the same teammate write the title, set the duration, lay the product shot, and leave you a sequence you can still open.” If the city is useful, it arrives as media in the project files and sits on a frame like anything else.
The standard we care about is cross-format collaboration that includes video. If a Shyne video ever becomes a ZIP you download and cannot reopen as frames, we have left that standard. The editable timeline and the rendered MP4 have different jobs: one is the source, the other is a delivery file.
A practical first film
Do not start with a manifesto. Start with fifteen seconds you would actually send.
Create a timed sequence with three frames. Frame one is the title and the mark. Frame two is the product or the chart. Frame three is the ask. Give them durations you can say out loud. Add one voice or one music bed from project files. Ask the teammate to write the three lines to a named audience. Correct the line that is trying too hard. Export MP4. Watch it on your phone. Fix the one frame that does not read at that size. Export again.
That is a complete video session. You did not open a second suite. You did not rebuild the brand. You did not lose the stills when the clock was added.
If you need a longer cut later, add frames. If you need a social square later, duplicate the sequence and change the frame size. The original remains. Identity is stable. Publishing can then read the exact sequence you selected.
Agents on the timeline
The teammate is not limited to copy. Timeline tools exist so an agent can inspect, place, trim, and check. Give it bounds. “Trim the second clip so it ends on the smile, do not touch audio, do not retitle frame one.” A vague “make it pop” on a timed sequence is how you get a new edit you did not ask to undo in pieces.
If the agent edited from a stale read, undo and make it re-read the playhead or the frame. The same stale-read failure that steals a heading will steal a cut. The fix is the same. Do not invent a video-only recovery ritual.
Review with people, not with a private movie
Share the project. A viewer can watch without being able to publish. An editor can move a title. You can sit in the same sequence and argue about whether the third frame needs another half second. That argument is cheaper on a live canvas than it is in an email with a 400MB attachment named v7-final-REAL.
When the edit is right, export it or post it to a connected channel. Publishing a film follows the same permission boundary as publishing a still: an agent-initiated send pauses for your approval, and the persisted release status is the delivery truth.
Markers, stills, and arguments
Markers exist so you can name a moment without writing a novel in the chat. “The smile,” “the logo hold,” “the cut that feels late” are better notes when they sit on the timeline. Ask the teammate to inspect a marker as a still. Argue about that still. Then change the cut.
A film review that only happens as a finished MP4 is a review that happens too late. Keep a still you can point at. Keep a frame you can select. The MP4 is for the people who cannot join the project. Everyone who can join should be looking at the sequence.
Keep video on the graph
The temptation will always be to flatten. Flattening feels like finishing. It is usually postponing the next change. Keep video on the graph until someone outside Shyne needs a file. Then compile. Then come back to the sequence, because they will ask for a different last line, and you should still be able to change a node.
Video belongs on the canvas because the canvas is already the document. Time is a property of a sequence. It is not a reason to leave.