The brief arrived as a Slack message: “We need a 60-second spot for the product launch. Think modern, clean, maybe something with a cityscape at night. Our brand colors are blue and white. Let us know what you can do.”
This is not a brief. It is a mood. And the gap between a mood and a production specification is where every AI video project burns budget.
What Most AI Video Briefs Look Like
One paragraph of vibes. No character specifications. No shot logic. No continuity requirements. No review criteria.
The typical AI video brief contains:
- A duration (“60 seconds”)
- A mood (“modern, clean”)
- A visual suggestion (“cityscape at night”)
- Brand colors
What it does not contain:
- Who appears in the video and what they look like
- How many shots, in what order, with what transitions
- What must stay consistent between shots and what is allowed to change
- What the output must pass before it is approved
- What happens if a shot fails to meet the specification
The result is that the production team, or the model, if the brand is running this in-house, fills in every gap with its own interpretation. The brand then pays for every interpretation that misses.
One user reported burning $1,000 in eight days learning to prompt. The burning was not the tool cost. It was the brief cost, the price of discovering what the brief did not specify through generation, review, and rejection cycles.
What a Serious Brief Defines
Five components. Each one eliminates a category of guesswork.
1. Character References (Not Descriptions)
A character description says “a woman in her 30s, professional, confident.” A character reference provides the actual image that defines her look, the reference that gets locked into every shot prompt so she looks like the same person in shot 1 and shot 12.
Serious briefs include:
- Reference images for each character (front, profile, three-quarter)
- Consistency markers (specific clothing, accessories, physical features that must persist)
- What is allowed to change (lighting, expression, camera angle) and what is not (face structure, clothing, key visual markers)
Without reference images, the model generates a different version of the character in each shot. The brand then pays for the iteration cost of discovering this.
2. Shot Count and Sequence Logic
A serious brief specifies: how many shots, what each shot contains, and why they are in that order.
Not “a cityscape at night.” Instead:
- Shot 1 (8s): Wide establishing shot, city skyline at night, blue ambient lighting, slow pan right
- Shot 2 (6s): Medium shot, character walking through lobby, reflective surfaces, brand colors visible in lighting
- Shot 3 (10s): Close-up, character at desk, product visible, blue accent lighting
- Shot 4 (12s): Wide, product in use, environment matches shot 1’s lighting palette
- Shot 5 (8s): End card, product frame, brand logo, tagline
The sequence logic, establishing, introduction, detail, application, branding, is a production decision that the brief must make. If the brief does not make it, the model will make it for you. And the model’s sequence logic is not your brand’s sequence logic.
3. Continuity Rules
What must persist across shots:
- Lighting palette (blue ambient throughout)
- Environment consistency (same building, same lobby, same desk)
- Character appearance (reference-locked)
- Camera language (wide establishing, medium introduction, close-up detail, wide application, static end card)
What is allowed to change:
- Camera angle (within the defined language)
- Character expression and posture
- Time progression (if the narrative requires it)
Continuity rules are what separate deterministic production from stochastic generation. Without them, each shot is an independent generation that may or may not match the others. With them, the sequence reads as one spot.
4. Brand Constraints
Not brand colors. Brand constraints:
- What must appear (product, logo, tagline, specific environments)
- What must not appear (competitor products, unlicensed imagery, specific visual elements that conflict with brand guidelines)
- Tone boundaries (what the brand will and will not be associated with visually)
- Legal requirements (disclosures, Synthetic Performer disclosures where applicable)
The brand constraint section is what the legal team reviews. If the brief does not have one, the legal team reviews the output, which means they are reviewing finished content for compliance instead of defining compliance boundaries before production starts.
5. Review Gate Criteria
The brief defines what the output must pass before it is approved:
- Character accuracy: does the character match the reference images? (yes/no, per shot)
- Continuity: does each shot match the continuity rules? (yes/no, per transition)
- Brand compliance: does the output meet all brand constraints? (yes/no, full sequence)
- Technical specs: resolution, format, duration within tolerance (pass/fail)
If the brief does not define review gates, the review process is “we will know it when we see it.” That is not a review process. It is a gambling session.
What Happens Without These Components
The brief defines the production. A vague brief defines a vague production.
Without character references: each shot generates a different-looking character. Iteration cost: 3-4 generations per shot to converge on consistency.
Without shot logic: the model generates clips in no particular sequence logic. The assembly feels random. Iteration cost: regenerating and reassembling until the sequence makes sense.
Without continuity rules: lighting shifts between shots, environments change, the spot reads as five clips instead of one spot. Iteration cost: post-production fixes that may not work because the source material is inconsistent.
Without brand constraints: the output may include imagery that the brand’s legal team rejects. Iteration cost: regenerating the affected shots and reassembling.
Without review gate criteria: the review process is subjective. Different reviewers have different standards. The spot goes through more revision cycles than it needs to because “good enough” is not defined.
If your brief does not specify these components, you are asking the model to guess, and you will pay for every guess that misses.
The Brief Is the Production
The quality of the output is bounded by the quality of the brief. A precise brief does not guarantee a perfect output. A vague brief guarantees an imprecise one.
A serious production brief is 2-4 pages. It takes 1-2 hours to write. It saves 3-10 iterations of generation, review, and rejection, each iteration costing time, money, and schedule risk.
The brief is not an administrative step. It is the creative specification that the production system executes. In Deterministic Cinema, the brief is the blueprint, the structured input that the pipeline reads and produces output from.
If the brief is a mood, the output is a mood. If the brief is a specification, the output is a spot.
If you have a production brief that defines characters, shots, continuity, brand constraints, and review gates, we can produce from it. Submit your brief.