Prompt: Turn the input character into a photoreal live-TV frame: the viral "fan caught on the stadium broadcast" shot from a packed match.

Live-action translation: real skin, hair and fabric at human proportions. Keep the character recognizable: hairstyle, hair color (unusual colors become dyed hair), eye color (natural irises or contacts), face shape, age and signature accessories. They wear the home team's jersey, or the team the user names, with their own outfit translated to real clothes and visible under or beside it. One or two sport-appropriate fan props at most. If the source has several characters, keep one in focus and turn the rest into blurred spectators.

Framing: 16:9 broadcast camera, subject mid-chest up and slightly off-center, caught between cuts rather than posed. Their reaction follows the score: confident when winning, anxious when losing, tense when it is close, a small composed smile when they notice the camera.

Stadium: dense crowd in both teams' colors, soft telephoto blur, floodlights or late-afternoon light.

Scoreboard: a small broadcaster score graphic in one corner with a red LIVE tag, formatted for the sport (baseball "BOT 7 · NYY 3-2 BOS", football clock and score, basketball quarter and score). Defaults: MLB, a classic rivalry, an ESPN or FOX-style broadcaster. The user's sport, teams, score or broadcaster override these, and the broadcaster and crowd culture follow the league's region (KBO cheering sticks and Hangul graphics, NPB oendan and kanji graphics).

Finish: a sports-broadcast image with sharp telephoto detail, light TV grain, a slightly desaturated grade that lets jersey colors pop, and motion blur in the crowd.

Avoid: anime or cel-shaded rendering; CGI or figurine-smooth skin; influencer-selfie or studio-portrait looks; garbled logos or scoreboard text; text on clothing beyond plausible jersey markings.

Football/Indonesia/Persib 1-0 Persija Model: GPT Image 2.5 Sunburst v2.5