Transparency centre · Zodiac beta
AI, astrology and synthetic voice
How readings are created and when you hear a synthetic voice.
Contents7 sections
1. When you interact with automation
Horoscopes and automated replies are produced by software systems, sometimes using AI models. The interface may present a synthetic voice, not a person speaking live. An explicit astrologer request or chart submitted for review has a separate workflow. Andreea Dinca’s name and support for the project do not mean she personally read every result. Individual review must be indicated only where an actual review record exists.
2. How readings are created
Astronomical data are calculated separately from birth information and the relevant period. Interpretive text may be produced by local rules or a model, according to configuration. The inspected local mode prepares future readings and syncs EN/RO pairs; it is not inherently a trained AI model. Actual models/costs are tracked in admin where providers report them. A configured model does not guarantee every result uses it: cache, fallback or pregenerated readings may apply. We do not claim provider models were specifically trained with Andreea without contractual/technical evidence.
3. Age and context
Birth date determines the life stage used in text; for young children, the message should be read by their caregiver. Optional context may help wording, but the system should not assume unknown partners, jobs or abilities. Automated checks reduce inappropriate wording without guaranteeing elimination of all errors. An interpretation must not be used as a fixed label about a child or a clinical assessment. Report age mismatches through feedback.
4. Questions and limits
You may seek an astrological perspective on love, work, parents, spiritual balance, travel or life choices. Requests unrelated to astrology, such as weather, sports scores or arithmetic, may be declined, as may diagnosis/treatment requests. An incorrect refusal may be sent for administrator review while you remain in the application. An automated answer is not the astrologer’s personal opinion. For human involvement, use the explicit option available to your plan.
5. What reaches the voice service
When voice is generated, the selected general or full text, language and necessary parameters are sent to the server and ElevenLabs. Greetings may include a first name. Existing audio may be reused without regeneration; local playback is not a new synthesis request. Listening speed does not itself require a new message generation. Voice authorisation, retention and provider text-use settings still need verification before commercial publication.
6. Correction and promise limits
AI and translations may produce incorrect or repetitive wording. Calculated charts depend on birth time/place precision and settings; human review must not be assumed until marked complete. You can report the exact reply, period and issue without publishing intimate information. We do not guarantee predictions, relationship outcomes, health or earnings. Professional wording must preserve the service’s symbolic nature. Assessment of applicable AI transparency duties and licences is required for the final version.
7. Andreea’s personal materials
The library distinguishes confirmed personal materials by Andreea Dincă from automatic app interpretations and other authors’ contributions. AI voice assistance for authentic materials in other languages is explicitly labelled; do not assume all materials are personally reviewed. Courses, meditations, rituals, workshops, seminars, ebooks and videos still in preparation are unavailable and are not a paid delivery promise. Future account benefits and demo plans will have individual conditions announced before availability.
Have a question?
For questions about the application or your data, write to us.
Sources and checks
Sources and checks
Sources explain the reviewed framework; they do not certify project settings or contracts. The legal framework was consulted on 8 October 2026; MDN technical sources on browser referrers on 10 October 2026. Romanian legal consolidation, applicability and final approval must be rechecked. Cloudflare/MDN metadata sources, GDPR Articles 5 and 7 and EDPB Guidelines 05/2020 on consent were also rechecked on 10 October 2026. Availability in this project account is separate from provider documentation. The ipapi.is API documentation, privacy policy and public DPA were consulted on 10 October 2026; provider statements and the account contract are distinct evidence.
- Company identity / Identitatea firmei
- GDPR — ANSPDCP (Arts. 5–9, 12–22, 25, 28, 30, 32–35, 44–49)
- EDPB — Data subject rights / Drepturi
- European Commission — Information for individuals
- Legea 506/2004, art. 4 — forma actuală de reverificat / recheck current version
- AI Act — Article 50 (applicability review required)
- OpenRouter — provider logging
- ElevenLabs — privacy
- Cloudflare — customer DPA
- MDN — browser referrer availability / Proveniența disponibilă în browser
- Cloudflare — trusted request metadata
- Cloudflare — conditional bot-management signals
- MDN — Screen colorDepth/pixelDepth
- MDN — basic userAgentData availability
- GDPR — EUR-Lex, Articles 5 and 7
- EDPB — Guidelines 05/2020 on consent
- MDN — Referrer-Policy
- ipapi.is — API documentation
- ipapi.is — Data Processing Agreement / Germany endpoint
- ipapi.is — Privacy policy (provider statements)