Home / Engine / Chart and Pipeline Engine

Affirmology Automated Pipeline - Architecture and Build Plan

Updated May 13, 2026 · Affirmology_AutomatedPipelineSpec_v1.md

Summary. Purpose: Outline what it would take to go from "hand-crafted demo for one person" to "any attendee enters their birth data and walks away with a personalized audio and accompanying methodology video."

Affirmology Automated Pipeline - Architecture and Build Plan

Purpose: Outline what it would take to go from "hand-crafted demo for one person" to "any attendee enters their birth data and walks away with a personalized audio and accompanying methodology video."

This is the moat. Anyone can write affirmations. Very few teams can ship a system that derives a personalized sacred audio from a birth chart in under ten minutes. The automation is the product. The script is just the artifact.


What the Pipeline Does, End to End

A user enters their birth date, time, and location. Ten to fifteen minutes later, they receive:

  1. A personalized audio (their "soul's song") rendered in ElevenLabs voice, layered over a Suno or library music bed.
  2. A short methodology video showing the chart-derived overlays in sync with the audio.
  3. A PDF or web page of the Script Map, so they can see the receipts.

At a live event, the same flow can run on a kiosk or registration page, with the deliverable arriving on their phone via QR code or email by the time they leave.


The Eight Pipeline Components

1. Chart Data Ingestion

Input: Name, birth date, birth time, birth location. Process: Geocode location, calculate ephemeris (planetary positions at birth time), derive Human Design design date (88 days before birth), calculate Gene Keys hexagrams from sun/earth positions, compute numerology from name and date. Tools: * Swiss Ephemeris (open source, the gold standard for astrology calculations). * HumanDesignAPI or a custom calculation layer on top of Swiss Ephemeris. * Gene Keys are derivable from Human Design gates with a mapping table. * Numerology is direct math, no library needed. Output: A structured JSON object with all relevant chart data.

2. Chart-to-Themes Mapping Engine

Input: The chart JSON from step 1. Process: Identify the user's Activation Sequence gates, Cross of Incarnation, Profile, defined and undefined centers, Sun / Moon / Rising, life path number, current pinnacle, current personal year. Output: A structured "personalization brief" similar to the AffirmationScriptBrief document we wrote for your chart, but generated programmatically. Note: This is where Affirmology's intellectual property lives. The mapping logic (which chart elements pull which language themes, which shadows pair with which gifts, how the narrative arc adapts to different cross configurations) is the actual product moat.

3. Script Generation

Input: Personalization brief. Process: LLM call (Anthropic Claude API recommended, given the spiritual nuance and consistency requirements) with a tightly constrained prompt that includes: * The personalization brief. * The 5-act narrative structure (Arrival, Descent, Standing, Reorientation, Return - or your final version of the arc). * Style constraints (first person, walking-appropriate, no em dashes, Sacral-friendly cadence). * The Script Map output requirement (so each line is tagged with its chart source for downstream visual overlay). Output: A structured script JSON, with each section tagged for both audio synthesis and visual mapping.

4. Audio Synthesis

Input: Script text from step 3. Process: ElevenLabs API call with the user-selected (or default) voice, stability and similarity parameters tuned for slow intimate delivery. Output: MP3 or WAV file of the spoken script. Cost reference: Roughly $0.30 per thousand characters at current pricing. A 925-word script runs around $1.50 to $2.00 per audio.

5. Music Selection or Generation

Input: Tone tags from the personalization brief (warm, devotional, theta-frequency, walking-tempo, etc). Process: Two paths. * Library path: Pre-generate a library of 30 to 50 Suno tracks at various lengths, moods, and tempos. Tag each. Match user's profile to closest fit. Cheap and reliable. * Generation path: If Suno opens its API (currently limited access), generate a custom track per user from the personalization tags. Higher cost and slower, but ultimate personalization. Recommendation: Start with library, move to generation when Suno API permits.

6. Audio Mixing

Input: Voice MP3 and music MP3. Process: Programmatic mix using FFmpeg or PyDub. Voice on top track, music underneath, ducked by approximately 12 to 18 decibels when voice is present, fades at intro and outro. Output: Final mixed MP3. Cost reference: Compute pennies per render.

7. Video Overlay Rendering

Input: Final audio file plus a JSON of script map overlays with timestamps (start time, end time, line 1 text, line 2 text, line 3 text per overlay). Process: Two recommended tools. * Remotion (React-based programmatic video). Build the template once in React. Each render takes the JSON, drops the right overlays at the right timestamps over the audio waveform, exports MP4. Modern, maintained, designed exactly for this. * FFmpeg with drawtext filters. More primitive but completely free and very fast. Good for MVP. Output: MP4 video. Cost reference: Compute, roughly $0.10 to $0.50 per render depending on resolution and length.

8. Delivery

Input: Final audio MP3 and final video MP4. Process: Upload to user-specific storage (S3, R2). Generate signed URLs. Email the user a link, or surface in a web dashboard. For events, generate a QR code that opens the deliverable on their phone. Output: A URL or QR code the user can use immediately.


Layer Tool Why
Backend Next.js (TypeScript) on Vercel Fast iteration, generous free tier, scales
Chart calculation Swiss Ephemeris (via npm wrapper) Industry standard, open source
LLM Anthropic Claude API (Sonnet for cost, Opus for quality on premium tier) Best for nuanced spiritual writing
Voice ElevenLabs API Already the standard you are using
Music Suno library (curated) + Suno API when accessible Two-tier approach
Mixing FFmpeg Free, battle-tested
Video rendering Remotion Programmatic React video
Storage Cloudflare R2 or AWS S3 Cheap audio and video hosting
Database Supabase (Postgres + auth + storage in one) Fastest path to MVP
Auth Clerk or Supabase Auth Drop-in user management

Build Timeline (Realistic)

Phase 1: Hand-Crafted Demo, validated process (this week, you are here). Manually run the workflow once for your chart. Confirm the methodology produces something worth automating. You have already done this.

Phase 2: Single-Chart MVP (2 to 4 weeks). Hard-code your chart, build the audio mixing and video rendering automation. The output is still "your audio," but every render after the script is locked happens by command. This is what lets you generate variations and iterations rapidly. Useful for fundraising demos.

Phase 3: Parameterized Pipeline (2 to 3 months). Add the chart calculation layer, the themes mapping engine, the script generation prompt with tested guardrails, and the user-facing form. By the end of this phase, anyone can enter a birth chart and receive a personalized audio. This is the demo you can show to investors and run at events.

Phase 4: Production Polish (3 to 6 months). Voice selection per user, multiple cross-of-incarnation arc variants, edge case handling (charts with rare configurations), user accounts, payment, premium tiers, the methodology video as a separate offering. This is the shippable product.

Phase 5: Scaling and Defensibility (6 to 12 months). Library of Suno tracks tagged by chart profile. A/B testing of arc structures by Profile and Cross. Multi-language. The Personal Sankalpa premium tier. The cohort offerings.

Total time to a real product: Six to nine months of focused work, assuming one strong full-stack engineer plus you. With a small team (engineer plus designer plus a contracted voice engineer for audio mastering), four to six months is achievable.


Unit Economics

Rough costs per personalized audio plus video:

Component Cost
Chart calculation Pennies (compute)
LLM script generation $0.10 to $0.50
ElevenLabs voice $1.50 to $2.50
Music (library) $0.00 (amortized)
Audio mixing Pennies
Video rendering $0.10 to $0.50
Storage and delivery Pennies
Total variable cost roughly $2 to $4 per piece

If your retail price for the personalized audio plus video is $49 to $99, gross margin per unit is 90 to 95 percent. Premium tiers ($199 to $499 for extended audio, multiple lengths, longer methodology video, follow-up sankalpa for next chapter) can be even higher margin.

This is a healthy unit economic profile for an investor conversation, especially compared to typical app or coaching businesses.


Live Event Variant

The full pipeline takes 5 to 10 minutes per render at MVP. At an event, that creates a queue. Three ways to handle it:

  1. Pre-event registration with birth data. Attendees submit data during registration. Audios render in the background. By the time they walk in, theirs is ready. QR code on their badge.
  2. Live kiosk with batched rendering. They enter data at a kiosk, get a number, return 15 minutes later. Coffee bar nearby helps.
  3. Hand-picked celebrity charts for stage demos. For an investor pitch, hand-craft the video for one or two well-known charts (Steve Jobs, Oprah, the investor in the room with permission) and play them. Higher production value per unit, lower automation requirement.

For investor demos in the near term, option 3 is by far the strongest path. For real events with paying attendees, option 1 is the right pattern.


What This Unlocks for the Business

Hand-crafted affirmation tracks are a service business with linear scaling. Each track requires a writer. Margins are tight, growth is bounded.

The automated pipeline is a software business with non-linear scaling. The same engineering effort that produces one audio produces ten thousand. Margins compound with volume. This is the difference between a coaching practice and a product company. It is also why this is the right thing to build for investor capital, because investor returns require non-linear scaling.

The methodology video output is the additional moat. Every audio Affirmology produces comes with visible receipts. That visibility is shareable, screenshot-able, and embeds the brand wherever the audio is consumed. It is also defensible. A competitor can copy "personalized affirmations from your chart." A competitor cannot easily copy "the system that proves it, in real time, with every line tagged to a specific gate or placement."

That is your patent-able moat, if you want to look at it that way. The Script Map is the product feature that makes the entire offering defensible.