Two capabilities, one engine. Get a brutally honest read on what your photos, bio, or full profile are actually doing — then let the engine generate the photographic moment and the written voice that would make the specific person you want stop scrolling.
Most dating tools tell you whether your photo is "good." That answers the wrong question. RizzBot answers two real ones: what is my current profile doing, and what would the right profile for my actual target look like.
Three rate modes. A stratified random sample of your target audience plus a fixed expert panel plus an optional dream-match panel of five top-tier vision models. Mr Rizz writes the cold read before they vote and the synthesis after. The final output is a predicted swipe rate computed from a deterministic formula, not a vibe.
You describe yourself, you describe your dream match, you pick marry or smash. The engine writes a calibrated photographic moment from your actual life and renders a photoreal reference image — not a portrait of you, a moodboard you can shoot your own version of. The bio writer picks three voice anchors that would resonate with your specific target and writes a bio in each.
Each rate mode runs the same panel architecture but evaluates a different artifact. Pick the one that matches what you want a read on.
One image. The panel sees only the photo and judges it cold. Use when you want to know if a single picture is working as a lead, an accent, or shouldn't be in the deck at all.
Output: predicted swipe rate (or SMG distribution), Mr Rizz's cold read, per-panelist verdicts with reasoning, MediaPipe face-read data feeding the panel.
Text only. The panel reads what you wrote with no visual frame. Use to test whether the words alone communicate what you want them to, before you pair them with photos.
Output: predicted hit-rate from bio alone, what the words signal, where the panel split on intent vs delivery.
Up to four photos and a bio together. The panel first sees the lead photo alone and renders a lead-only verdict, then sees the full profile and renders a final verdict. The delta is the empirical impact of your bio and accent photos.
Output: two rates (lead-only and full-profile), per-panelist verdict shifts, swap-lead recommendation if a different photo would perform better.
The panel mirrors the audience you're trying to reach. Same architecture, different pool and different verdict vocabulary.
Stratified random sample of women from the pool (young / mid / older) plus a fixed panel of dating experts. Each renders SWIPE: RIGHT or SWIPE: LEFT. Mr Rizz computes a predicted swipe rate, narrates the cold read, and synthesizes the panel into a final integrated verdict.
Stratified random sample of men plus a fixed panel of male experts. Each renders SMASH, MARRY, or GHOST. The trinary captures something a binary swipe can't — a profile can be SMASH-heavy with no MARRYs (high desire, low respect) or MARRY-heavy with no SMASHs (long-term signal but low immediate spark). Mr Rizz reads the distribution.
Most photo generators produce generic attractive people. The dream engine produces the specific photographic moment from your actual life that would land on the specific match you want — and the bio voice that would make him save the profile and read it twice.
You describe yourself in plain language. You describe your dream match. The engine writes a calibrated photographic prompt — pulling shared-life cues from MATCH ("wants someone to read in the park with") and intersecting them with your actual life. An age-band aesthetic routes the visual treatment (Gen Z anti-aesthetic / Millennial Kinfolk-romantic / Gen X editorial). A side-conditional eye-contact script handles the moment of recognition. A Sharp post-processing pass adds real film grain for the Gen X register. You get a photoreal reference image to shoot your own version of.
The bio writer reads your self-description and your dream-match description, identifies three radically different authorial voices that would resonate with this specific target, then writes a bio in each voice as a transformation of who you are. Toggle marry or smash — different registers, different bios. The voices are intentionally diverse so the writer can feel real range and pick the one that fits her mood.
The recognition look in a marry photo is "HE LIGHTS HER UP" — happiness is the ground state, expressed in the register that fits her temperament (shy contained / open warm / quiet complicity / playful caught-laughing) crossed with his attraction triggers. The smash version is charged frank desire — pouty and sultry are fine. The eye-contact script only fires when the chosen moment produces eye contact; absorbed moments bypass it.
When MATCH names shared-life activities — "wants to read in the park," "travel together," "wake up with," "cook together" — those are HIS explicit fantasy-cues, not throwaway language. The engine extracts them, intersects with HER actual life, and weights them heavily as scene candidates. If her life supports the activity, lift it directly. If not, find the closest equivalent from her world that triggers the same fantasy register.
Across regenerations for the same SELF+MATCH pair, the engine tracks scene, location, micro-moment type, and aesthetic medium — and forces a meaningfully different facet on each subsequent render. Direct-engagement moments and absorbed-unaware moments alternate. Different bands rotate. Different MATCH-cued fantasies cycle. You get a coherent set, not four variations of the same energy.
Mr Rizz is the analyst who frames every evaluation, reads your submission cold before anyone else does, and writes the final synthesis after the panel votes. He sees what the panel saw and what they said. The panelists never see him.
Before any panelist is consulted, Mr Rizz looks at the submission alone. He reads the values being communicated or undermined, the role the submission is playing on the app, what's working and what's failing structurally. This is the uncontaminated baseline you'll compare everything else against.
After the panel votes, Mr Rizz reads every panelist's verdict and reasoning. He integrates the consensus, names the split where the panel disagreed, and writes the final word. He's allowed to overrule the cold read if the panel showed him something he missed — and to flag when the panel converged on a misread.
Mr Rizz isn't your friend. His job is accuracy, not encouragement. The framework anchors on values communicated by the gestalt — masculine on the men's side, feminine on the women's side — and on whether your submission lands clearly or muddles its own signal.
68 voices on the men's side, 54 on the women's side. Every rate pulls a stratified random sample across age bands — young, mid, older — so you get spectrum coverage but never the same exact lineup twice. The point isn't that one specific archetype loved or hated your profile. The point is whether the pattern of reads across the spectrum tells a story.
Stratified sampling across age bands ensures you always hear from young, mid-range, and older voices. Within each band, archetypes vary widely — from validation-seekers to career peers, from spiritual-seekers to transactional pragmatists. Each carries its own value framework, language, and threshold for what's worth swiping on. These run on faster smaller models for cost and latency at scale.
A fixed panel of expert voices renders structural reads independent of audience preference. Dating coach, behavioral scientist, matchmaker, PUA, plus a critic role. They tell you what role the photo plays on the app, whether the gestalt holds together, and what's getting filtered out before any swipe even happens. Each expert runs on a different top-tier vision model so they don't echo each other.
Describe the specific person you're trying to reach and the system generates five distinct variants of that archetype, each running on a different top-tier vision model. They read you alongside the random pool. The divergence between random-panel and dream-panel tells you whether your profile is hitting your actual target or only resonating broadly.
The dream-match panel runs on five genuinely distinct top-tier vision models — no two share a brain, so the voices stay independent:
The final number isn't an opinion. It's a computed product of three measured things, normalized against a realistic baseline.
For full profiles, each panelist renders TWO verdicts: one after seeing only the lead photo, one after seeing the full profile. The two computed rates — lead-only and full — let you measure the empirical impact of your bio and accent photos. If the rates are equal, the bio added nothing. If full is higher, the bio is lifting. If full is lower, the bio is costing matches.
The experts also indicate whether the current lead photo is the right pick or whether one of the accents would perform better as the lead. A consensus shift gets surfaced as a swap-lead recommendation — a 30-second change that can move your rate more than rewriting your bio.
When the dream-match panel runs, the system reports the random-panel rate and the dream-panel rate separately. A profile can rate well to the random audience but poorly to your actual target, or vice versa. That divergence is often more useful than either number alone.
Pick the language for your evaluation and Mr Rizz, the panel, and the experts all speak it natively — not translated word-for-word from English, but in the language's actual rhythm, idioms, and verdict conventions.
RizzBot is a simulated panel reading a static submission, plus a generation engine writing a calibrated reference. It's better than asking your friends — but it's not a focus group, not a clinical study, and not a fortune teller.