How HeyBooker finds, scores and applies.
The same rules the product runs on, written down. Scoring weights, eligibility gates, the scam shield and the apply methods, so you can tell when a score is right and when it isn’t.
1. Sources and enrichment
Casting calls are collected throughout the day from casting platforms, agency and studio boards, and the Instagram and TikTok accounts where many castings are posted first. Duplicates that appear on several sites are merged into one listing with every source attached.
Each listing is read by AI to extract what a booker would: the disciplines involved (modeling, acting, commercial, background and extras, runway, influencer/UGC, voiceover), project type, gender and age range sought, ethnicity notes, location or remote, pay (paid, TFP, unpaid, deferred, unknown), union status, deadline and how to apply. When a field is not stated and the model has to infer it, that inference is flagged as such, and it is never allowed to disqualify you (see gates).
2. The match score
Every casting gets a 0–100 score against your profile. The factors and their maximum points:
| Factor | Points | What it measures |
|---|---|---|
| Location | 25 | Same city, same state or region, or remote. The single largest factor. |
| Semantic fit | 15 | How close the casting’s text is to your whole profile (an embedding comparison, computed server-side). |
| Casting type | 10 | The listing’s disciplines against the work you said you want. |
| Keywords | 10 | Your own search terms found in the listing. |
| Age | 10 | Inside the stated range. |
| Gender | 10 | Explicit match, or unspecified (never penalised when the casting is open). |
| Ethnicity | 8 | Only when the casting states a requirement and you chose to list yours. |
| Pay | 5 | Paid over TFP over unknown over unpaid. |
| Union | 4 | Non-union or “both” favoured, since most talent is non-union. |
| Freshness | 3 | Recently posted and not about to close. |
Bands: 75+ strong, 55–74 good, 35–54 possible, under 35 low. Your feedback (save, pass, “not my look”, “too far”, “pay too low”) sharpens future rankings.
3. Hard gates and caps
Some things are not a matter of points. A casting is marked ineligible, not just down-ranked, when it is closed, its deadline has passed, it explicitly seeks a different gender, or you fall outside its explicitly stated age range (with a year of slack either side). These gates fire only on what the casting states outright; an inferred age or gender never disqualifies anyone.
Structural mismatches are capped rather than hidden: an off-type casting (a discipline you didn’t ask for), union-only work when you are non-union, or a language requirement you don’t list, can score no higher than 45. It stays in your feed, labelled, and never masquerades as a good match. This came from calibrating against castings a real model hand-graded: off-type listings that scored 70 on location and demographics alone were never a fit.
4. The scam shield
Every listing is screened for the patterns that define casting fraud: fees to apply or to be “considered”, portfolios, classes or comp cards you must buy as a condition of the job, wire-transfer and overpayment scripts, and contact details that don’t match the poster. Rejected listings never reach any user. Borderline ones are shown with a visible caution. The public pages on this site show only clean passes, and never castings for minors.
5. Applying
For each casting HeyBooker resolves the best way in, in this order of preference: a direct submission on a connected casting platform (made under your own account, after you confirm); an email to the casting director with your comp card attached, sent from HeyBooker on your behalf; a web form, opened with a copy of your answers ready to paste; or an Instagram or TikTok DM, with the draft on your clipboard and the profile open. Drafts are written in your voice from your profile and the listing, in the language the casting was written in, and are always shown to you first. Every application is tracked, and you record what came back: heard back, booked or no answer.
6. Your profile and photos
The onboarding chat builds your profile the way a booker would take it down: look, stats, home base, languages, the work you want and the markets you’ll travel to. Photos you upload are tagged so the right ones accompany the right applications, and the comp-card generator lays out a card from them. Profile and photos are shared only where you point an application. Delete your account and everything goes with it.
7. What this is not
HeyBooker recommends opportunities to you. It does not rank or screen candidates for casting directors or employers, and no caster sees your profile unless you apply. Scores are recommendations for your eyes; they are not decisions made about you.
When the score is right, and when it isn’t.
What is a good match score?
75 and up is a strong match, 55–74 good, 35–54 possible, below 35 low. Your feed can be filtered by a minimum score, and Top Picks shows the strongest matches first.
Why does a casting I like score low?
Usually location: it is the largest single factor. Structural mismatches (a different discipline than you asked for, union-only work when you are non-union, a language requirement you don’t list) cap the score at 45 so they stay visible but never pose as a good lead. Open the casting and the breakdown says exactly which factor cost the points.
Can a wrong guess hide a casting from me?
No. When the source didn’t state an age range or gender and the enrichment inferred one, that inferred value never disqualifies you; it only nudges the score. Hard gates fire only on what the casting itself says.
How fresh is the catalog?
Sources are read throughout the day; the feed refreshes from the catalog every half hour, and the daily digest emails you when new castings match, at most once a day.
Where does the AI run, and on what?
Google Gemini reads casting posts (public content) to extract role details and screen for scams, tags the photos you upload, builds your profile from the onboarding chat and drafts your applications. Under our API terms none of it is used to train third-party models. Every draft is reviewed by you before it is sent.