Prototype · Shams TV · 2026

AI as the cast, not the crew.

A hybrid human-AI show for a news channel. Real presenter, real reporter, and the AI models you already know — ChatGPT 5.1, Claude Opus 4.5, Gemini 3, DeepSeek 2 — played as the cast.

A five-minute pilot concept. DeepSeek 2 gets a section of its own. Made with AI, not by AI (more on that below).

The premise.

News + AI is a minefield — especially once you're on a current affairs channel where trust is the whole product. The usual answer is: keep AI out of sight. This is the opposite: put it on screen. Give each frontier model a role. Let the audience watch how X's Grok handles a story differently to Google's cautious Gemini. Turn the differences into content.

The point isn't to replace journalists. It's to build public AI literacy — through showing, not telling. Capabilities, limits, biases. Entertaining, informative, honest about what each model is and isn't.

Underneath all of that the older question still applies: can we make something compelling enough that people want to watch?

Don't compare AI output to conventional content.

When a new medium arrives, we reach for the old one to measure it. Early photographs were compared to paintings. The comparison was natural and also wrong — photography is a different medium.

AI video is a different medium from traditional video. Trying to make it replace live video is a bad use of the tool. The more interesting question is what new things become possible.

The low-hanging fruit is magic — fantasy worlds, stunts, science fiction, animation — things that were previously either impossible or punishingly expensive. But bigger than that: AI is to video what Photoshop was to image. A new instrument for precise manipulation across every medium.

Slop is made by people, not by AI.

The generic tsunami of bland, soulless content — and the darker waves of dis- and mis-information — aren't the tool's fault. Humans pick what to make with it.

How big is your prompt?

The most annoying, inauthentic use of AI is AI content generated without thought or effort. Even if you generate a whole novel, its value reflects the work you put in.

Scale is easy with AI. Originality and authenticity aren't. How much of the output is really yours? There is a visible difference between two minutes of input and two weeks.

Made with AI, not by AI.

The video above wasn't made BY AI. It was made WITH the help of AI. Every shot had to be imagined, every character brought to life, every movement and voice worked scene by scene. Every element demanded a specific tool:

• Visual style + mood boardsMidjourney
• Character creationMidjourney
• Character placementNano Banana (v1)
• VoicesEleven Labs
• Movement + cameraSeedance, Kling, Hailuo
• Lip syncOmni Human, Lipsync Pro
• Human acting → character movementWan
• Editing actor appearance / locationNano Banana, Seedream, Wan
• Dialogue + voice conversionEleven Labs
• Rap + sound effectsSuno, Mirelo
• UpscalingTopaz Labs
• Final edit + subtitlesFinal Cut Pro, DaVinci Resolve

Two weeks of hard work to produce five minutes. That's thirty seconds of finished video a day. You can't make a weekly show at that pace.

A whole new production crew.

If this concept ever becomes a real show, it would need a crew as big as the budget allows. Even with consistency libraries and repeatable templates in place, the people operating this chain of tools are the new production roles. Prompt designers, model directors, continuity artists for AI characters, ethics reviewers for what goes on air.

Your talent and expertise are more important than ever — not less.