News
Fuji TV Sets Up an AI Sandbox and Aims to Use Generative AI Across Its Dramas by October 2027: What Is Claimed
ยท RenderBob team
Fuji TV announced an AI Center and AI Sandbox on 5 November, and aims to use generative AI across its dramas by October 2027. The 70-80% workload cut is its own figure.
Japan's Fuji TV announced on 5 November an AI Center and a testing site for generative AI in drama. It says AI cut workload by up to 70-80% in certain segments. The figures are the broadcaster's own.
What was announced
Automaton, on 6 November, reports the following. A Korean outlet, SBS, also covered it, but we could not open that article.
- Structure. An AI Center inside a new Digital Strategy Headquarters runs AI Sandbox, a testing site at Fuji TV's Wangan Studio. Directors, producers and technical staff can work with outside AI producers there and decide whether generative AI fits a project. It has been in use since September.
- Timeline. From October 2026, at least one live-action drama per season is to be an "AI-enhanced" project made through the Sandbox. By October 2027, Fuji TV aims to have a system for using generative AI across all its running series, depending on each project's needs. Automaton's headline says "by the end of 2027" while the article body says October 2027, and the article does not explain the difference.
- Rationale. Fuji TV cites changing filming conditions and the need to fully realise each project's worldview and the director's vision.
The example
The latest season of the long-running medical drama Emergency Room 24 Hours already uses generative AI. Per the article, AI is used in pre-production to visualise the director's intended look, for image processing in medical scenes and to depict accident scenes. It cites an AI-generated explosion over an AI-generated aerial view of a fictional city in the first episode.
The 70-80% figure
Fuji TV says that since September, workload in certain video-production segments using generative AI has fallen by up to 70-80% against traditional methods. The article stresses that this applies to those segments only, not to overall workload or budget. We have no method, sample or independent check.
What is not settled
Automaton does not address actors' likeness, consent or the data used to train models. A weekly film-and-TV roundup that cites SBS says Fuji TV bars training AI on actors' faces without permission. We could not confirm that. A Taiwanese outlet's headline says Japanese online users reacted negatively and saw cost-cutting as the motive. We did not read it.
What to ask a broadcaster that adds an AI lab
- Which tasks use AI, and which do not?
- What does "up to 70-80%" cover, and against what baseline?
- What are the rules on likeness and training data?
- Who approves AI output, and how are changes tracked?
- How are staff roles affected?
- What is disclosed to viewers?
What to watch
Fuji TV's own release, the first "AI-enhanced" drama under the Sandbox, any statement on actors' likeness, and whether other Japanese networks publish similar targets.
RenderBob 3 and this story
Generated clips get their own timeline lane and the model is chosen per clip, so a team can test AI on one shot before committing a whole episode.
More from the blog
- Gods Don't Give Gifts, an AI-Assisted Animated Feature, Opens in US Theatres on 4 December: What Fable Says About the Making
Fable says about 30 people made Gods Don't Give Gifts in nine months. The AI-assisted feature opens in limited US theatres on 4 December. No budget is disclosed.
- Korean Filmmakers Warn AI Films Could Cost More Than the 35mm Era: The Numbers From Taiwan's Creative Content Fest
At Taiwan's Creative Content Fest, Korean AI-film producers said token prices and re-renders could push AI features above 35mm costs. The figures are theirs.