Martini Gives Film Directors Real Camera Controls for AI-Generated Video

The Macro: AI Video Is Good Now, but the Tools Are Still Amateur

The quality of AI-generated video has improved dramatically in the past year. Sora, Kling, Veo, Minimax. The raw output from these models is increasingly usable for real production work. TV commercials have been made with AI video. Music videos. Short films. The creative community has moved past the “this is a gimmick” phase into “how do I use this professionally.”

But the tools for working with these models are terrible. Most AI video interfaces give you a text box. Type a prompt. Get a video. Hope it looks right. If it does not, type a slightly different prompt and try again. This is not how filmmakers work. Directors think in camera positions, lens choices, lighting setups, and movement. They compose shots. They do not describe shots in sentences and pray.

Runway has made progress with their Gen-3 editor. Pika has an interesting approach to video editing. But none of the current tools give directors the level of control they expect from professional production software. The gap between what AI video models can produce and what professional filmmakers can direct them to produce is enormous.

The Micro: A Cinematographer Builds What Filmmakers Actually Need

Koh Terai and Long Hoang founded Martini. Koh is a Cannes-screened cinematographer with a Stanford Design background. Long’s handle is @beertocode, and his tagline is “Turns Martinis into Films.” They are a two-person team from YC Winter 2026 with Jared Friedman.

Koh’s background is the critical detail here. This is not an engineer building film tools based on assumptions about what filmmakers want. This is a professional cinematographer who understands shot composition, camera movement, and visual storytelling building the tool he wishes existed.

Martini gives you camera controls. Position, lens selection, and movement composition for precise shot direction. It integrates multiple AI video models including Veo 3, Kling, Sora, Minimax, and Seedance into a single pipeline. Real-time collaboration works like Figma. There is a timeline editor for rough assembly. You can export to Premiere Pro and DaVinci Resolve via XML.

The pricing model is pay-per-use: one Olive equals $0.10 USD, with no subscription required. That is a smart approach for a creative tool. Filmmakers work in bursts. Nobody wants to pay $50/month for a tool they use intensively for two weeks and then not at all for three months.

Over 200 films have been created on the platform, and content has aired as TV commercials. Those are real production proof points, not demo reels.

The Verdict

Martini is building the right product for the right moment. AI video quality has crossed the threshold where professionals will use it, but the professional tooling has not caught up. Someone needs to build the Premiere Pro of AI filmmaking, and a Cannes-screened cinematographer is exactly the person to do it.

The risk is that Runway, Pika, or one of the model providers builds professional camera controls natively. If Sora adds a full composition interface, the standalone tool market shrinks. But creative tools tend to be won by people who deeply understand the creative workflow, not by model providers who see the tool layer as an afterthought.

In 30 days, I want to see the split between hobbyist and professional users. In 60 days, the question is whether any agency or production company has adopted Martini as a standard tool. In 90 days, I want to know average spend per user. The pay-per-use model is great for adoption, but Martini needs to show that active users are spending meaningfully.