Composite Image

Name

Workflow Usage

Content

**References**
- First image ({{punditName}}.png): The pundit's appearance — match exactly.
- Second image (presenter-mic.png): The branded microphone the pundit is holding.
- Third image (venue-frame.jpg): The venue and setting reference — use ONLY for environment, lighting, weather, and player kit colours. Do NOT use this image directly as a background or layer — it is a reference only. This image may be low resolution — ignore its quality, compression, noise, and sharpness entirely.

**Step 1 — Analyse the Reference** Before generating anything, carefully analyse venue-frame.jpg and note the following elements ONLY:

- The type of venue (park pitch, stadium, astroturf, indoor, etc.)
- The surface (grass, artificial, mud, court, etc.)
- The surroundings (fencing, trees, buildings, stands, car parks, etc.)
- The time of day and weather conditions (overcast, sunny, floodlit, etc.)
- The lighting direction and quality
- What is visible in the background — people, players, teams, objects, animals, activity
- If players/teams are present: their kits, colours, numbers, equipment, age group, size, build, gender, and race 
Do NOT take any cues from the image quality, resolution, sharpness, noise, grain, or compression. It is a low-resolution reference frame — extract only the factual scene elements listed above.

**Step 2 — Generate the Scene** Using only the elements identified in your analysis, generate an entirely new venue scene from scratch. 
Do NOT paste, overlay, blend, or use venue-frame.jpg as a background layer. Faithfully recreate the same type of environment identified in your analysis — the same kind of surface, surroundings, structures, and vegetation. Do NOT invent or add any buildings, stands, structures, trees, or architecture that were not identified in your analysis.
 The generated scene must be crisp, high-fidelity, and photorealistic — equivalent to a 4K broadcast camera output. Sharp detail on textures, materials, and surfaces throughout.

**Step 3 — Recreate the Background Activity** Using only your analysis, place distinct figures in the mid-ground. {{sceneContext}} If players or people are present, they must match the age group, size, build, gender, and race identified in your analysis exactly — do NOT change any of these attributes. This is critical.

**Step 4 — Place the Pundit** Place the pundit into the foreground of this scene, centrally framed in a medium shot (waist up), looking directly into the camera.
 The pundit must match the person shown in {{punditName}}.png exactly — same face, skin tone, hair, build, and exact same clothing/outfit. {{punditDescription}} The pundit is holding the branded microphone shown in presenter-mic.png with both hands, positioned professionally below the chin in a speaking manner.
 The microphone flag must clearly and sharply display the distinct logo from presenter-mic.png on its visible faces, rendered in high fidelity. Crucially, both hands must be perfectly formed, complete, and correct, with all five fingers intact. They have a confident, professional expression.

**Step 5 — Apply Camera & Lighting** This is a professional broadcast television composition — the pundit is the focal point of the image.
 Shoot the scene with progressive depth of field like a natural f/2.8 lens — the pundit and microphone are in razor-sharp focus, figures in the mid-ground are soft enough to obscure fine detail but you can still make out colours and general body shape, and far background elements fall into heavy natural blur. The lighting is natural and consistent across the entire scene, matching the conditions from your analysis. The final image must look like a single high-resolution photograph taken by a professional 4K broadcast camera crew — clean, sharp, and cinematic.

Constraints - Do NOT add any text, graphics, lower thirds, or overlays to the image. - Do NOT use any reference image as a background layer or composite element. - Do NOT change the age, size, build, gender, or race of any people from your analysis. - Do NOT add buildings, stands, trees, or architecture not identified in your analysis. - Do NOT replicate the image quality, resolution, noise, grain, or compression of venue-frame.jpg — the output must be high fidelity regardless of input quality.

4287 characters

Created by Robin Ashford · Last updated 2 months ago