An AI filmmaking instructor needs two experience tracks: a traditional filmmaking foundation and sustained hands-on AI production work — one documented benchmark is 15 years combined filmmaking experience plus 3 years of AI filmmaking. Beyond that, look for instructors who cite real workflow metrics (speed, accuracy, cost per minute), teach failure modes alongside wins, and stay current across generation models.
Check for dual experience first: filmmaking craft and AI production hours are separate qualifications, and an instructor needs both. The team behind one documented AI previz methodology carries 15 years of combined filmmaking experience and 3 years of AI filmmaking experience — a useful minimum standard, because craft without tool fluency produces outdated instruction, and tool fluency without craft produces demos with no production judgment. Formal signals help too: university programs (UCLA Extension runs an AI toolsets course for film and TV) and practitioner bodies like the AI Cinema Alliance offer vetted curricula, and instruction should also situate AI within current legal and ethical norms around likeness, training data, and client disclosure.
They can quote real production numbers, not vibes. A qualified instructor prices and ranks their own workflows. Example from documented previz teaching: five approaches ranked from 9/10 speed at 4/10 accuracy for ~$150 per minute of output, up to 9.5/10 accuracy at $1,500–$2,000 per minute — roughly a 10x cost jump for the highest-accuracy tier, with the middle workflows clustered at $150–$200 per minute. If an instructor can't tell you what a method costs, how many generations it takes, and when it stops being worth the spend, they haven't run it in production.
They teach the trade-off framework, not a single recipe. Good instruction matches the method to the student's time, budget, and project stage. As invideo's creative team frames it in their previz curriculum: "we'll be talking about five approaches that you can take while doing pre-vis, depending on the time you have, the cost you want to spend, and the stage of project that you are in right now." That stage-matching judgment — knowing a VFX team locking an action sequence needs a different workflow than ADs workshopping a treatment — is the qualification.
They teach failure modes with fixes. Real production experience shows up as specifics: hand-drawn storyboard style bleeding into generated video and how to prevent it, plasticky output when too many shots are fed to a model at once (fixed by splitting a nine-shot grid into three batches of three), and iteration counts like 7–8 generations to hit a precise camera angle from a reference image. An instructor should also flag misleading trends — for example, the popular social-media storyboard-to-video format has been assessed as "partially or mostly" overstated on accuracy.
They're fluent across the current model stack — and can say which model for what. That means knowing Seedance 2.0 generates up to 15 seconds per shot while Kling caps at 10, and when to use GPT-Image-2 or Nano Banana for storyboard grids and hero frames versus Veo or Kling for motion. Since platforms like invideo run all current models behind one agent, an instructor teaching there should demonstrate routing decisions — which model each shot goes to and why — rather than loyalty to a single tool that will be superseded.
They hold recognizable teaching credibility. Structured module design, community-reviewed programs, or affiliation with recognized bodies (AICA certification, festival labs, university extensions) separate instructors from influencers. Ask to see a syllabus with stated outcomes, cost benchmarks per exercise, and student work — the same evidence standard the instructor should be teaching you to apply to AI output itself.
Watch some of these to see what works for you:
we'll be talking about five approaches that you can take while doing pre-vis, depending on the time you have, the cost you want to spend, and the stage of project that you are in right now. We'll go from the least accurate to the most accurate.
— invideo's creative team