Produce video and images with generative AI tools

Sixty-second AI-generated product spot with a reusable prompt library

Invent a small D2C brand (a chai, skincare or sneaker label) and write a 60-second script with one recurring presenter character and one hero product. Storyboard it into 8–10 shots, generate a key still per shot, and choose an image-to-video tool from the toolkit. Start with a short test shot and check your available credits before generating the full sequence. Add a voiceover and an avatar intro using tools you can access. Log every prompt, seed, reference and retake in a prompt library so the same character and product look identical from the first shot to the last.

A 60-second script and a one-page brand sheet (name, colours, product shape, presenter description) you write yourself, with or without an LLM — no footage, no client

1. Write the script and brand sheet, then storyboard 8–10 shots with a shot-level prompt for each (subject, action, camera, lens, light, mood) · ~4h
2. Lock the presenter and product: generate reference stills, pick seeds, and start the prompt library with working and rejected prompts · ~6h
3. Test one shot in your chosen image-to-video tool, then generate the sequence within your available credits; switch tools only if needed to keep character and product consistent · ~8h
4. Voice the script and produce an avatar intro with your chosen tools; check lip-sync and pacing against the script · ~3h
5. Run an artifact QA pass on every clip (hands, text, geometry, physics), regenerate the failures and write the retake notes · ~2.5h

• 8–10 finished clips exist in which the presenter's face, costume and the product's shape and colours match across every shot, with a contact sheet proving it
• The prompt library lists, for every shot, the model, prompt, seed or reference image, number of retakes, and why each rejected take was rejected
• A voiceover of the full script and an avatar segment of at least 10 seconds exist with no visible lip-sync drift
• A QA log names at least five AI artifacts you caught (garbled text, extra fingers, warped packaging, etc.) and how each was fixed or retaken

Evidence a recruiter can check

• A contact sheet of all 8–10 shots showing the same presenter and product from first shot to last, with the reference stills they were locked from
• The prompt library as a spreadsheet or markdown table — model, prompt, seed/reference, retake count and rejection reason per shot
• A QA log with before/after frames for at least five artifacts you caught and fixed
• The raw clips, voiceover and avatar segment, with a note on the tool, model and access plan used for each

Produced a 60-second AI-generated product spot with one consistent presenter and product — backed by a prompt library that logged the tools, models, seeds, references and retakes.