How Workshop Handles Real Production

Workshop is a motion graphics compiler. You write a .ws file, and the compiler handles typography, color, timing, animation physics, and layout — producing deterministic, brand-safe video from structured intent.

This guide walks through the real production scenarios that prompt-based tools can't handle, and how Workshop's architecture solves each one.

Intent In, Design Out

Workshop doesn't generate code from a prompt and hope for the best. You write a .ws file — a structured, human-readable script — and the compiler executes a deterministic pipeline that makes every design decision for you.

Film "Q3 Update"
  mood: confident
  theme: corporate-blue

Scene "Headline" [5s]
  show "Revenue grew 40%"
    size: hero
    enter: rise
  beat
  show "Our strongest quarter yet."
    size: title
    color: accent

This isn't a prompt. It's a specification. The same .ws file produces byte-identical output every time. You can version-control it, diff it, review it, approve it, and automate it.

The Compiler Encodes What Designers Know

Professional motion designers make hundreds of micro-decisions per video that audiences never consciously notice. Workshop's compiler encodes this knowledge:

Psycholinguistic pacing. Hold durations are calculated from syllable counts and cognitive load. A mood: calm video holds text 40% longer than mood: urgent — because the reading-speed model accounts for the difference between relaxed and pressured viewing.
Physics-based motion. Animations are driven by spring dynamics (mass, tension, friction), not keyframes. Every movement has physical weight.
Typographic hierarchy. The type scale, weight, spacing, and color are resolved from the mood profile automatically. You write size: hero and the compiler calculates the exact pixel size, weight, letter-spacing, and line-height.
Constraint-based layout. The layout engine calculates mathematically correct positions for every element based on hierarchy and safe margins.

Workshop's mood system encodes restraint as a first-class property. A mood: confident video uses clean cuts, subtle rises, and long holds. A mood: elegant video favors negative space and deliberate pacing. Fifty parameters per mood profile, all tuned to a specific emotional register — including knowing when not to animate.

Production Scenarios

Brand consistency across every asset

The problem: Brand guidelines specify exact colors, approved fonts, minimum clearance, and a motion vocabulary. Every video needs to match, regardless of who made it or when.

Workshop's answer: ThemePacks.

A ThemePack is a versioned JSON bundle that encodes a brand's entire visual identity — palette, typography, spacing, timing overrides, motion preferences. Register it once. Every .ws file rendered through it is on-brand by construction.

Legal approves the ThemePack once. Every video thereafter complies automatically. When the brand updates its color palette, update the ThemePack — every .ws file re-renders with the new palette without touching a single script.

Revisions without rework

The problem: The client changes the headline. Then the CTA. Then asks to swap scenes. In a prompt-based tool, you re-generate the entire video and pray the rest stays the same.

Workshop's answer: edit the .ws file.

Changing a headline is changing a string. Swapping scenes is cutting and pasting two blocks. The rest of the video is guaranteed identical because the compiler is deterministic. No drift, no re-rolling the dice on the parts you already approved.

diff

  Scene "Headline"
-   show "We're launching something new"
+   show "Meet the all-new Platform X"
      size: hero
      enter: rise

One line changed. One line different in the output. Everything else: byte-identical.

Matching previous campaign style

The problem: This quarter's video needs to "feel like" last quarter's. In a prompt-based tool, you can't reproduce a feel.

Workshop's answer: same mood, same theme, same templates.

If last quarter used mood: confident with theme: corporate-blue and the quarterly-report template, this quarter uses the exact same settings with different content. The visual identity is structural, not accidental.

Multi-format delivery

The problem: You need 16:9 for the website, 9:16 for Stories, 1:1 for LinkedIn. Each format has different safe areas and text scaling.

Workshop's answer: recompile with a different aspect ratio.

The layout engine recalculates positions, the type system rescales, and safe margins adjust — all automatically. One source file, as many outputs as you need.

Templates for repeatable content

The problem: You produce weekly updates, monthly reports, quarterly reviews. Each one should have the same structure, just different content.

Workshop's answer: define the template once, fill in variables.

define template stat-slide [6s]
  background: dark
  show "{heading}"
    size: display
  beat
  show number "{value}"
    size: hero
    count-from: 0
    label: "{label}"
    color: accent

Scene "Revenue" from stat-slide
  heading: "Revenue"
  value: "$4.2M"
  label: "Total Revenue"

Approved once. Reused every week with new data. The structure never drifts.

Subtitles and localization

The problem: The video needs subtitles in English, French, and Spanish. Timing has to adjust because French runs longer.

Workshop's answer: the timing engine is language-aware.

Workshop's syllable counter has language-specific strategies for English, Spanish, French, German, Japanese, Chinese, and Korean. Hold durations recalculate automatically per language.

Delivery at scale

The problem: You need a personalized version for every customer, SKU, or sales rep. Hundreds or thousands of variants.

Workshop's answer: API + batch rendering.

27 REST endpoints including batch rendering with job status polling. Feed in a CSV of variables, get back rendered videos. Multi-tenant org isolation means each team operates in their own namespace.

Where AI Helps — And Where It Doesn't

Workshop isn't anti-AI. It uses AI where it's genuinely useful and enforces determinism where it's required.

Where AI helps: Workshop's AI copilot takes a plain-English prompt and generates a complete .ws script. Mention a URL — "make a launch video for acme.com" — and the brand resolver extracts colors from the page automatically. The LLM handles creative reasoning: what to say, how to structure the narrative, which mood fits.

Where determinism takes over: The compiler validates every AI-generated script. If the LLM produces invalid syntax, the error trace feeds back automatically for self-correction — 100% syntactically correct output. Once the .ws file exists, everything downstream is deterministic. No hallucination, no drift.

AI for reasoning. A compiler for execution. Neither alone is sufficient.

Quick Start

The simplest Workshop video is three lines:

Scene "Hello"
  show "Hello, World"

The compiler fills in: confident mood, title size, primary color, fade entrance, 16:9 aspect, calculated timing. From there, every property is overridable — you control exactly as much or as little as you want.

To see it in action, follow the Quickstart → and compile your first video.

For the full language reference, see the Workshop Language Reference. For API integration, see the API Reference.