Opportunity brief
Date of Event is 15 October 2026
Pixel Prompt is a fast-paced game of visual description and AI generation. One team member (the Describer) is shown a secret target image for about 2 minutes to observe and memorize key details. They must then convey those details to their teammate (the Prompter). Using only those verbal clues, the Prompter constructs a detailed text prompt to generate an image via an AI tool that matches the original photo as closely as possible. The team whose generated image comes closest to the original wins.
Duration: 15-30 minutes (adjust based on the event)
Guidelines:
- Team Structure: Every team must consist of exactly 2 members—Member 1 (Describer) and Member 2 (Prompter).
- Observation Phase: Member 1 will be shown the target image for 2 minutes.
- Member 2 must look away or remain blindfolded during that time.
- Communication Phase: Once the time is up, Member 1 has 4 minutes to verbally describe the image to Member 2. No gestures, drawing, or showing written notes are allowed.
- Prompting Phase: Member 2 will write a prompt into the designated AI image generator within 6 minutes to render the image.
Rules and Regulations:
- Generation Limit: Only one prompt attempt (or a maximum of two, if specified before the event) per round is allowed per team.
- No Direct Assistance: Member 1 cannot touch the keyboard or assist Member 2 in writing the actual text prompt.
- Disqualification: Any team caught taking photos, using smartphones during the round, or looking at the target image out of turn will be disqualified immediately.
Evaluation criteria:
- Subject & Core Layout (10 points ): the actual representation of the image generated.
- Chromatic Fidelity & Color Accuracy (30 points ): measures how closely the generated image reproduces the colors, color relationships, intensity, and overall color mood of the original target image.
- Dual Image Comparative Exactness (20 points ): evaluates the side-by-side similarity between the original target image and the AI-generated image. The judge compares both images directly and determines how accurately the generated image reproduces the original.
- Prompt Syntax & Descriptive Precision (20 points ): evaluates how effectively the Prompter converts the Describer's verbal information into a clear, structured, and specific AI prompt. This directly aligns with the event's criterion of evaluating the effectiveness, descriptive detail, and efficiency of the prompt.
- Background & Environmental Context (20 points ): evaluates how accurately the generated image recreates the surrounding setting and environmental elements of the original image—not just the main subject. This is important because the original Pixel Prompt criteria specifically include similarity in key elements and overall composition.
Note: The detailed scoring system and weightage for each criterion will be announced by the event coordinators.