Photorealistic AI Sports Sequences

The project revolves around building a library of broadcast-quality sports footage generated entirely with AI. Every second of video must look as if it was captured by a real stadium camera: natural lighting, authentic player motion, believable crowds, crisp uniforms and gear—all indistinguishable from live action.

Scope
• Produce short to mid-length clips (around 15–30 seconds each) that can slot straight into highlight edits, promos or training breakdowns.
• Sports are still open; feel free to suggest which ones your current AI pipeline already renders convincingly.
• Photorealism is non-negotiable. Stylised looks or obvious artefacts will be rejected.
• Any workflow is welcome—Stable Diffusion, Runway Gen-2, Pika Labs, custom diffusion models, Unreal Engine, or a hybrid—as long as you hit the realism bar.

Deliverables
1. Final 4K video for each approved storyboard (ProRes HQ or lossless PNG sequence).
2. Source assets: prompts, model settings, project files or Unreal scenes so lightning-fast tweaks are possible later.
3. One iterative revision pass per clip to polish timing, camera angles or colour grade.

Acceptance Criteria
• Accurate anatomy, equipment and physics throughout the clip.
• No warping, jitter, frame skips or texture flickers when viewed frame-by-frame.
• Consistent lighting and colour across every frame.
• Ready-to-use audio track optional, but crowd ambience or commentary is a plus.

When replying, include at least one link to a photorealistic AI video you have produced—sports content preferred—to show you can already meet this visual benchmark.