Fuji TV Builds an AI Sandbox for Every Drama, Writers Not Yet Heard

Fuji Television said on November 5 that it aims to have a system in place by October 2027 to use generative AI in any drama it produces, wherever the work or the scene calls for it ^[2].

That turns an experiment into a standing method at a major Japanese broadcaster. Fuji says it makes 16 terrestrial prime-time serial dramas a year ^[2]. The plan reaches all of them in principle.

In principle is the phrase that matters. Fuji does not say every scene or every show will be made with AI. It says AI will be available as a production method, to be used when it suits the project ^[1][2].

A sandbox in Odaiba

The network has set up an AI Center inside its Digital Strategy Headquarters and an AI Sandbox at its Wangan Studio, where it tests generative AI on live-action drama ^[2]. The Sandbox began operating in September. Fuji's drama and technical staff work there alongside outside AI producers and designers and the AI-related company ANDENT ^[2][3][4].

The schedule is concrete. From the October 2026 programming period, Fuji plans to designate at least one drama every season as an AI-enhanced project ^[2][3][4]. The group-wide ambition is larger. Fuji says it wants to extend the capability to other video companies in its group and aims for AI capability across the group by fiscal 2029 ^[4].

The examples Fuji gives are modest and visual. Generative AI is already in use on several dramas, including the current season of Emergency Room 24 Hours, for pre-production visualization, image processing in medical scenes and depictions of accident scenes that would be hard to film ^[1][2][3].

In the first episode of that series, the aerial view of the landscape around an explosion was generated with AI, and smoke was composited onto it ^[1][2][3][4].

Fuji's producers describe the tools as a way to show a director's idea before the cameras roll. Producer Ryotaro Adachi said the team uses AI-generated images and video before filming to share the director's intended scene and to work out how live action and image processing might fit together ^[2]. Fuji drama and film production executive Daisuke Nagashima said the tools could help with scenes that are hard to shoot because of location, weather, legal or safety limits ^[2].

The Sandbox's technical lead, Shinya Masaki, said the aim is not to make everything with AI. He said it should be combined with established methods such as VFX, with the best tool chosen for each production. He added that the project is meant to build in-house knowledge and widen visual expression as well as raise efficiency ^[2][3].

The number and its fine print

The figure that will travel furthest is 70% to 80%. Fuji reports that generative AI cut work by as much as that in certain video-production processes, compared with traditional methods ^[1][2][3][4].

Fuji itself says the number describes specific processes. It is not an average across a production and it is not a cut to a show's total budget ^[1][2][3][4]. A figure like that tells you a tool can be very good at one narrow job. It does not tell you what happens to the job around it.

What the announcement leaves out

Here is the paper's view. Fuji's announcement does not show a network trying to replace writers, actors or directors. Nothing in the examples given so far touches scripts, dialogue or performances. A synthetic aerial view behind a fireball is closer to a matte painting than to an AI actor.

But the announcement is still a bet. A broadcaster does not build a center, a sandbox and a four-year timetable to save a few hours on explosions. It has told its staff and its industry that the efficiency gains are real enough to organize around, and it has said so before telling them what happens to the hours that shrink.

Those questions are open. Fuji has not said which paid tasks will get smaller, whether crew sizes or working hours will change, or who answers for AI-generated material.

It has not said which models it uses or what training material sits behind them. Nor has it said how it will credit human contributors or tell viewers when AI was used. And it has not said whether generative AI will ever reach scripts, dialogue or likenesses of performers, or stay confined to images.

Nor has it said what will count as AI use in a drama in 2027. A single composited background and a substantial part of a production are very different things. A label that covers both will tell viewers little.

The actors have already drawn a line

In Japan the performers have started writing rules. On October 15, three weeks before Fuji's announcement, the Japan Actors Union published guidelines on voice and AI. They call for consent before an actor's voice is used to train AI or to create AI audio that replicates it, for human performers to stay the priority, and for compensation to reflect any use of a voice ^[5].

The guidelines also say performers who refuse, or later withdraw, consent should not be disadvantaged in casting, auditions, employment conditions or other treatment ^[5]. That last clause shows where the union expects pressure to come from. It is not the technology itself. It is the quiet penalty for saying no.

Nothing in the sources says Fuji's drama plan involves cloned voices or likenesses, and the union's guidelines are not a response to it. They show the standard that any later move toward performers' voices or faces would be measured against.

What to watch

The first real evidence arrives with the first dramas Fuji designates as AI-enhanced from October 2026. Watch what the credits say, what the production staff say, and whether the network publishes how it reached its 70% to 80% figure. Fuji's writers, actors, directors and crews have not yet been heard in any of this, and that silence is the most important gap in the story.