The Natalis method
Calculated precisely. Interpreted transparently. Composed beautifully.
The chart engine, interpretation rules, image generation, quality control, and physical production are separate steps. That separation protects both astrological accuracy and the artwork you choose.
1. The calculation is deterministic
Swiss Ephemeris calculates planetary positions, angles, houses, and aspects under a pinned Natalis configuration. The language and image models never calculate or invent placements. Every project stores the engine version, settings, time certainty, effective house system, and a hash of the calculated result.
Natalis launches with the Western tropical zodiac and Placidus houses, using an explicit Porphyry fallback where Placidus cannot be calculated reliably. Unknown birth time removes the Ascendant, Midheaven, and houses instead of guessing them.
2. Interpretation has a named lineage
Astrology contains multiple schools. Natalis does not flatten them into “astrologers say.” Each interpretation rule belongs to a declared tradition, source proposition, and studio decision. Our launch canon begins with:
- Dane Rudhyar and The Astrology of Personality, for a humanistic whole-chart orientation.
- Liz Greene and the psychological-astrology tradition represented by works including Saturn: A New Look at an Old Devil, for symbolic depth without keyword reduction.
- Demetra George and Ancient Astrology in Theory and Practice, for disciplined attention to planetary condition and traditional technique.
- Robert Hand and Horoscope Symbols, for understanding astrological components and bringing them into synthesis.
- Deborah Houlding and The Houses: Temples of the Sky, for the history and philosophy behind house meanings.
These astrologers do not endorse Natalis. We name them so customers can see which work informed a rule, where traditions differ, and where Natalis studio authorship begins.
3. The written summary is grounded sentence by sentence
A rules engine retrieves only propositions supported by the verified chart and the active source canon. It produces a claim ledger before any prose is written. An approved Azure-hosted narrative model may organize those allowed claims into concise, poetic, or explanatory language, but a validator checks that every sentence points back to a chart fact, customer choice, approved source proposition, or explicit Natalis authorship.
Optional name, pronoun, and identity rendering happens in Natalis code rather than being inferred by the model. Identity never changes the chart calculation or automatically selects a palette, body, motif, or gendered visual trope.
4. The art brief is bounded
Chart structure and customer choices become a typed art brief. It contains only the visual decisions the model needs: art world, orientation, palette family, density, geometry, approved motifs, literalness, atmosphere, crop-safe regions, and derived structural emphasis. It does not contain a customer’s name, email, raw birth data, raw quiz prose, or full chart JSON.
5. Two exact concepts are revealed before payment
Two independently generated concepts pass file, dimension, crop, artifact, brief-adherence, and visual-difference checks. The customer compares private watermarked versions in the browser. If both are valid, no curator delay is inserted. Human review is reserved for exception cases that automated quality control cannot confidently approve.
One bounded refinement can adjust art-direction controls without changing the underlying chart. Once selected, the artwork hash is frozen. The print master derives from that asset—not a new creative generation.
6. The physical result must match
Each poster, canvas, and framed-canvas variant has a deterministic print recipe for pixels, crop, wrap, color profile, sharpening, and checksum. Natalis must inspect physical samples before a route becomes an approved launch product. The exact-preview promise is complete only when the wall object preserves the composition the customer selected.