NEWS · September 2026

Shotline: my friends built the music video studio the rest of us kept faking

Shotline: my friends built the music video studio the rest of us kept faking

Full disclosure before anything else: Aidan Yagu is a friend of mine. I am not a neutral party here. What I am is fairly hard to impress by tools, because most creative software that launches with a dark landing page and a lowercase manifesto turns out to be a thin coat of paint over a single API call.

One more bit of honesty before the praise starts: I sweated. There is a learning curve here and I climbed every step of it the hard way. Three months ago I was downloading stock footage and dropping it into CapCut. I had never heard of Minimax or Wan 3, and if you had said the word node to me I would have assumed you meant something in a network cabinet. So when I tell you the first sessions were work, that is coming from a man who started at absolute zero. The curve is real. It is also a curve and not a wall, and somewhere in the first week the thing that looked like a cockpit starts to feel like a desk.

Shotline is a different animal. It is the first tool out of Third AI Studios, the two-person outfit Aidan runs with SÆra Nova, and it exists for a very unglamorous reason: they were making music videos together across two screens and the workflow was held together with exports, file names and hope.

If you have ever made an AI music video, you know the workflow I mean. A subscription for images. A different one for video. A third for upscaling. A folder called final_v3_REAL that contains four files, none of which are final. Somewhere in there you were also supposed to be creative.

One board, four rooms

Shotline puts the whole pipeline on one canvas and splits it into four spaces that hand work to each other. Nothing gets exported to move between them.

The Director is where you decide what the video looks like before you generate a single frame. You pick a director's look, then a real camera rig underneath it: body, focal length, lens character, aperture, film stock. That optical language then rides along on every prompt, which is the whole point. It is the difference between forty shots that share a vision and forty shots that share a folder. Cast and sets are locked here too, so the same face stays the same face, and you can save your own looks and pull them into later projects.

The Grid is a node canvas, and it is where the pipeline stops being invisible. You wire characters, styles, references and prompts into generators, chain an image into a video, branch off, compare two versions, keep the one that works. Every model is a node and outputs feed the next input, which means the thing you built at 2am is still readable at noon the next day. There are starter templates for the moves everyone makes anyway: coverage, restyle, upscale, atmosphere, zine.

Build is the bench where you actually press the button. One prompt slab, a rail of models with their real capabilities listed, and here is the part I like most: the exact price of the run, before you commit, compared across every provider that carries that model. It also shows speed and reliability numbers learned from the whole fleet, so you know whether the cheap route is also the route that times out. Every run gets logged with its recipe, so you can rerun the exact thing three weeks later instead of trying to remember what you typed.

Signal is the room that made me sit up. Your song is the timeline. You drop a track, the lyrics get aligned to the waveform for you, and every line takes a shot. The cut is built on parallel shotlines: one main lane that the player reads, side lanes holding your alternates, and you drag shots between them while the audio re-cuts itself underneath. Each shot quietly collects its frames, its audio and its takes into a brief a model can actually read. One button sends that brief to the Grid, and the finished take comes back to the timeline on its own.

That loop, timeline to generator and back again without you playing courier, is the bit no other tool in this space has bothered to build.

Behind all four there is the Library, where every generation, upload and reference files itself automatically. Browse it, bulk select it, send it somewhere, run it again.

The multiplayer part

Shotline was built by two people directing one video on two screens, so collaboration is not something bolted on for the enterprise pricing page. It is the foundation.

Live cursors, real co-editing, nodes changing under your eyes as your co-director changes them, and nothing to merge afterwards. You invite people by handle or link as editor or viewer. Every board carries its own chat thread with presence and unread state, so the conversation happens where the work is instead of in a Discord DM you will never find again. Billing stays with whoever pressed generate, which sounds like a small detail until you have co-directed something with a friend and had to work out who owes who.

There is also a preset system where looks, styles and atmospheres become codes anyone can import in one click. Expect that to turn into a small economy of shared house styles, which is exactly what should happen.

Thirty-nine models, none of them resold

Here is the commercial angle, and it is refreshingly blunt. Shotline does not sell you generations. You bring your own provider keys and every run is billed by the provider at their own listed rate. No markup, no plan floor, no credit bucket that expires while you are busy at your day job.

The catalog currently runs to thirty-nine models across four providers: twenty for video (Seedance, Kling, Hailuo, Wan, LTX, Grok Imagine and friends), eight for image, six for audio including stem separation and ElevenLabs, and five finishing tools for sound design, matting, depth and reframing. The compare card prices your shot on every carrier before you commit and routes to the cheapest automatically. The site even says that on the rare month where somebody else's bundle would beat this setup, the card will tell you so. That kind of line is easy to write and awkward to live with, and they wrote it anyway.

Your keys also stay yours, which means the credits on them still work on everything else you build. You are not renting a walled garden.

The GPU lane, and the nicest idea in the whole thing

Studio Plus adds a second route: rent a graphics card on your own account and Shotline queues your shots onto it. A card charges by the hour instead of by the clip, and that quietly changes how you work. Once the card is warm, take twenty costs almost nothing on top of take one, so iteration stops being a financial decision every single time. It also unlocks inputs the hosted endpoints simply do not sell, like following a reference clip frame for frame, or building a shot against a sheet of your own cast. Restyle, inpaint, outpaint, clean plate, relight and upscale all run there too, against footage you already have.

And then there is the gift line. A warm card is a card you can lend. Open your line and a friend can move a queued shot onto it: their job, your GPU, the result goes back to them, and your key never leaves your account. No money changes hands. There is a leaderboard ranking hosts by how much they have actually rendered for other people, so the ones keeping the lights on sit at the top. In a scene that mostly optimises for follower counts, ranking people by generosity is a genuinely good idea.

What you get, and what Plus adds

Both tiers include the four rooms, the shared canvas, cast and sets, the live cost compare, the free in-browser tools, song mastering in the browser, ViZ for music-reactive visuals, and an MCP connection so you can drive the studio from your own AI assistant.

Studio Plus is for when the work becomes properly yours: authoring your own styles, director looks, atmospheres and camera presets, your own prompt bible steering the enhancer, a wardrobe system for many outfits on one consistent face, the ViZ+ composer with 31 effects, plus the GPU lane and the gift line.

Early access is open, and it has conditions

They are picking a handful of artists to finish V1 with them. Not a waitlist, a crew. Two conditions, and they mean both: anything you publish from early access carries the short Made in Shotline clip they provide, and you actually report the bugs you hit using the built-in reporter. As they put it, ten artists who ship beat a thousand who lurk.

Your work stays yours. The badge just says where it was forged.

If you make music videos with AI and you are currently paying four subscriptions to do badly what this does in one place, go apply. And if you get in, report your bugs properly. I have to see this man at some point.

Apply for early access

Share

Comments

Loading…

I'm not built to behave and neither should you!