Hands Demonstrative Sessions on AI Tools - Day 5

Summary

The fifth-day session covered practical image generation in ChatGPT and introduced prompt engineering for educational and professional use.

Image-generation demonstrations

The instructor demonstrated using ChatGPT’s Create image feature to make:

• Personalized greetings: Attach a photo and specify the occasion, recipient, message, senders, and aspect ratio. Refine the result with follow-up instructions such as “make it more colorful,” then download or copy it.
• Invitations: Provide event details, including the organizer, date, time, venue, resource person, and invited-by information. Optionally attach a person’s photo or an institution’s logo. The instructor demonstrated changing the layout from horizontal to vertical and refining its style.
• Campus banners: Generate banners for themes such as anti-drug awareness and a plastic-free campus, with specified wording and aspect ratios.

The instructor also demonstrated altering a photograph to depict a person in a new situation. This capability should be used cautiously: obtain consent, avoid defamatory or harmful depictions, and do not circulate manipulated images maliciously.

Prompt engineering: concepts and guidance

Prompting was described as the input or language used to communicate a task to an AI model. Clear, specific prompts can improve relevance and accuracy compared with generic requests. In education, prompts should be tailored to the syllabus, learner level, and intended audience. Generated material should be reviewed and edited rather than used uncritically.

A prompt may be a question, instruction, scenario, example-based request, or code-related request. The instructor highlighted these common elements:

• Role: The expertise or perspective the AI should take.
• Task/action: What it should produce or do.
• Context and constraints: Background, level, language, length, or limits.
• Output format: Such as slides, bullets, tables, or an essay.
• Audience/target: Who the result is for.

The instructor emphasized iterative refinement: review the output, clarify or revise instructions, and add constraints until it meets the need. For ongoing work, continue in the same chat rather than starting a new one each time; earlier context can help keep later outputs consistent.

CRAFT model and activities

The instructor introduced CRAFT as a prompt-writing structure:

• C — Context
• R — Role
• A — Action or attribute
• F — Format
• T — Target

A worked example requested a Tamil PowerPoint on prompting styles, specifying the presenter’s context, the AI’s role, the task, the slide count and format, and the intended audience.

Practical activities and assignments:

1. Create a personalized greeting.
2. Design an event invitation.
3. Make a campus-awareness banner.
4. Experiment with image alteration while observing consent and ethical limits.
5. Write 10 prompts using the CRAFT model, covering different uses such as lecture notes and image generation.

The instructor said the day’s assessment would include 10 multiple-choice questions and 10 drag-and-drop questions. Participants were also asked to create a Zoom virtual background using the workshop flyer and their photograph. Guidance on adding it in Zoom was to follow in the next session.

Prompt types and further prompt models, including zero-shot and few-shot approaches, were introduced as topics for a later session rather than covered in depth.

Main speakers and sources

• Primary speaker: The instructor leading the AI-tools demonstrations and prompt-engineering lesson; not clearly identified by name in the subtitles.
• Other sources: A technical team assisting with the online session and participants responding to demonstrations and asking occasional questions.