# AI Tool Landscape — June 2026 Teaching Notes **Check this list before the first class and lightly update during the term.** Tool names change, features move, models get retired (Sora's consumer app was retired in April 2026 — proof that nothing here is permanent), and pricing shifts. Capability *categories* are stable; product names are not. **The instructor rule:** Never teach one tool as permanent. Teach the decision pattern: **problem → input → output → risk → verification.** --- ## Quick Reference: What We Actually Use Each Week | Weeks | Primary tools | Backup if unavailable | |---|---|---| | 1–3 | ChatGPT or Claude (free tier OK) | Either one; paper fallback exists | | 4 | Firefly/Photoshop generative, ChatGPT Images, Midjourney (demo) | Free-tier image gen, instructor demo + critique track | | 5–6 | Chat model + image model + one video tool (demo only) | Prepared clips/screenshots | | 7–8 | Chat model + Figma (existing student accounts) | Paper wireframes + chat model | | 9 | Chat model + Blender (free) | Prepared scripts, screenshots | | 10 | Transcription tool + chat model with image input | Phone voice memo + manual captions | | 11–12 | Claude Code or Codex (instructor demo), free chat for specs | Spec-writing track needs no tools | | 13–15 | Chat model | Paper + peer review | One paid frontier-model subscription (~€20–25/month) covers a student for the whole course. Free tiers + pairing cover everyone else. --- ## Core Text and Multimodal Assistants The "chat" models. All are multimodal (text, images in/out, file uploads); differences are in voice, reasoning depth, and ecosystem. - **ChatGPT (OpenAI — GPT-5.5 family as of April 2026).** Broad default. Strong at briefs, critique, research planning, coding, and image generation (ChatGPT Images 2.0 leads on text rendering inside images — useful for design mockups). Free tier exists with rate limits; Plus tier is the workshop's recommended baseline. - **Claude (Anthropic — Opus 4.8 / Fable 5 family).** Strong writing voice, long documents, careful structured thinking, critique, and coding. Currently the strongest coding model — relevant to Weeks 11–12. Free tier exists; Pro recommended. - **Gemini (Google — 3.x family).** Strong Google ecosystem integration (Docs, Drive, NotebookLM), long context, good multimodal/video understanding. Free tier generous. NotebookLM is excellent for research-source grounding (Week 3). - **Perplexity-style research tools.** Source-linked answers; better than raw chat for "find me real references" — but links still require verification. Week 3 staple. **Classroom stance:** students pick ONE chat model and learn it properly rather than tool-hopping. Demos alternate between ChatGPT and Claude so both communities see their tool. ## Image Generation and Editing - **Adobe Firefly + Photoshop generative features.** Safest classroom fit: trained on licensed/Adobe Stock content, lives inside production tools students already use, and NCAD students typically have Creative Cloud. Generative Fill/Expand are the Week 4 workhorses. - **ChatGPT Images / Gemini image generation.** Strong instruction-following and text rendering; good for layout roughs, previs, and iteration inside a conversation. - **Midjourney (v8).** Strongest stylized aesthetics; subscription-only, Discord/web. Use as instructor demo unless students already subscribe. Discuss its training-data controversies honestly — it's the canonical ethics case. - **Krea / Leonardo / Reve-style tools.** Fast iteration, real-time canvas tools, layout-aware editing (Reve 2.0). Free tiers shift constantly — verify before teaching. - **Upscalers (Topaz, Magnific-style).** Mention in Week 4/10 as finishing utilities; demo only. **Rights note for all image tools:** policy on commercial use, training-data sourcing, and indemnification varies by tool and tier. Firefly is the conservative recommendation for client-facing work; this is itself a teachable decision pattern (Week 4, Week 7). ## Video Generation and AI-Assisted Video - **Google Veo 3.x.** Current quality leader for text/image-to-video with audio. Access via Gemini tiers/Flow; limits change often — verify. - **Runway (Gen-4 family).** The most production-oriented suite: motion control, references, plus practical post tools (green screen, inpainting). Runway Academy is good teaching material (Week 5). - **Kling / Luma / Pika.** Fast iteration, image-to-video, effects. Free credits fluctuate. Sora 2's consumer app was retired April 2026 — use as the "tools die" case study (Week 15). - **Classroom reality:** video generation is demo-territory, not lab-territory: credits are expensive, queues are slow, and the pedagogy (previs, boards, shot planning) works better with image + text tools anyway. Weeks 5–6 are built accordingly. ## Audio, Voice, and Music - **Whisper-class transcription** (built into many editors; standalone apps; Descript). Captions, transcripts, edit-by-text. Week 10 workhorse — free options exist. - **ElevenLabs-style voice synthesis.** Temp VO, accessibility reads, scratch narration. **Voice consent is a hard rule** — no cloning real people without documented permission (Week 10 ethics block). - **Suno/Udio-style music generation.** Mention only: rights situation contested (ongoing label litigation); scratch tracks in private work only, never client deliverables. Flag as "verify legal state before any real use." ## 3D and Spatial - **Blender + AI-written Python.** The reliable, free, genuinely useful combo: scene scaffolds, repetitive setup, material/render checklists (Week 9 core). - **Meshy / Tripo-style mesh generators.** Image/text-to-3D. Outputs need retopology — that's the lesson, not a flaw to hide (Week 9's "Topology Reality Check"). - **Gaussian splat / NeRF capture tools (Luma, Polycam).** Phone-based capture for reference and previs; fun, accessible, free tiers. ## Coding and Agentic Tools - **Claude Code (Anthropic) and Codex (OpenAI).** Terminal/IDE agents that read and edit real project files, run commands, and build small tools. Weeks 11–12 core demo tools. Instructor account is enough; students write specs and watch builds, or use free chat models to generate single-file tools. - **Cursor / Windsurf-style AI editors.** For students already coding; mention, don't require. - **Lovable / Bolt / v0-style app builders.** Chat-to-deployed-app. Fast prototypes; students must learn to inspect and revise the output, not just accept it (Week 11 caution block). ## Research and Organization - **NotebookLM-style grounded notebooks.** Upload sources, get cited answers/audio summaries. The "RAG you can show" demo (Weeks 3, 12). - **Built-in deep-research modes** (ChatGPT/Gemini/Claude). Long cited reports; good for Week 3 and Week 13 (job/market research). Always paired with verification habits. --- ## Account and Access Policy **Recommended student kit (≈€20–25/month):** - ONE of: ChatGPT Plus or Claude Pro. That covers Weeks 1–15. - Optional for coding-curious students: Claude Code / Codex access (often included in the above tiers at limited rates — verify current bundling). **The free-tier path (fully supported):** - Free ChatGPT/Claude/Gemini covers most chat-based labs (rate limits = built-in iteration discipline). - Image labs: Firefly via NCAD Creative Cloud where available; otherwise free tiers + pairing. - Pairs/teams: one account per pair is always acceptable; labs are designed for it. - Every lab has a zero-generation track (planning/critique/documentation) requiring at most a free text model. **Institutional notes:** - Check whether NCAD's Adobe licence includes Firefly generative credits (it usually does; quota varies). - Students under 18: not expected at NCAD, but tool ToS minimum ages vary — flag if running outreach versions. - **EU/Irish context:** GDPR applies to anything personal students upload; the EU AI Act's transparency obligations (synthetic-media labelling) are phasing in — disclosure habits taught here are about to be law, not just ethics. Useful framing for admin conversations. **Data hygiene rules (post in classroom, repeat in Week 1 and Week 12):** 1. No client work, classmate artwork, or personal data of others into any tool without permission. 2. Assume anything uploaded may be retained — check each tool's training-data opt-out and use it. 3. Turn off chat-history training where the option exists; prefer business/edu tiers for sensitive work. 4. Generated content of real people (face/voice) requires documented consent. Always. --- ## When This Document Goes Stale It will. Re-verify each term: - [ ] Model names/versions in the Core Assistants section - [ ] Free-tier limits (image counts, message caps) - [ ] Video tool access routes and credit pricing - [ ] Adobe/NCAD licence coverage - [ ] Music-generation legal status - [ ] EU AI Act implementation dates relevant to disclosure teaching - [ ] Each weekly page's "Verify before teaching" resource flags (see `PRODUCTION_NOTES.md`)