Hotcheck

Analyzes uploaded photos through an undisclosed vision model to generate a numerical 'hotness rating' by evaluating four distinct dimensions: facial attractiveness, body attractiveness, style assessment, and photo quality. The system processes each image for approximately 30 seconds server-side, returning a blended composite score without per-dimension breakdowns. Architecture appears to use a cloud-based inference pipeline (hosted on Vercel) that extracts visual features and applies a proprietary scoring function, though the underlying model identity, training data, and exact scoring methodology remain undocumented.

Enables side-by-side analysis of two photos to determine which has higher viral potential by running both images through the attractiveness-scoring pipeline and returning a ranked comparison with mode-specific insights. The comparison mode costs 2 credits (equivalent to Pro mode pricing) and outputs a direct ranking statement ('Photo A works better') plus contextual reasoning. This capability abstracts away individual scores and presents a relative judgment, reducing cognitive load for users deciding between two options.

Generates text-based insights about photo attractiveness in three configurable modes: standard 'Quick Score' (basic summary), 'Pro Mode' (additional exclusive insights), and 'No Sugarcoating' (harsher, more critical tone). Each mode has different credit costs (1, 2, and 2 credits respectively) and output verbosity. The system appears to use conditional prompt engineering or separate model fine-tuning to vary tone and depth, allowing users to choose between encouraging feedback and blunt critique. A bundle mode combines Pro + No Sugarcoating for 3 credits, offering both detailed and harsh perspectives.

Implements a proprietary credit system to control access and monetize analysis operations. Users receive a limited free credit allocation (quantity undocumented) and can purchase additional credits in three tiers: Starter (5 credits for $12.99), Pro (12 credits for $24.99), and Max (25 credits for $34.99). Each analysis mode consumes 1-3 credits: Quick Score (1), Pro Mode (2), No Sugarcoating (2), or bundle (3). The system tracks per-user credit balance and enforces hard paywall when credits are exhausted. Purchases are one-time (no subscription), and credits do not expire (persistence model undocumented).

Provides new users with a limited free credit allocation to test the core attractiveness-scoring capability before requiring payment. The exact quantity of free credits is not disclosed in available documentation, nor are the conditions for credit replenishment, expiration, or reset. Users must create an account to access free credits, establishing a sign-in barrier that enables tracking and potential future upselling. The free tier appears designed as a conversion funnel: users experience the tool's core value proposition (single-image scoring) at no cost, then encounter a paywall when attempting higher-value modes (Pro, No Sugarcoating) or exhausting their allocation.

Processes uploaded images on Vercel-hosted backend infrastructure, extracting visual features (face, body, style, quality) and computing attractiveness scores via an undisclosed vision model. The analysis pipeline introduces approximately 30 seconds of latency per image, suggesting either complex feature extraction, model inference, or both. No client-side processing is mentioned, indicating all computation occurs server-side, which centralizes model access but introduces network round-trip delays. The architecture does not support batch processing or concurrent multi-image analysis — each image requires a separate 30-second request.

Claims to predict social media virality based on facial attractiveness, body attractiveness, style, and photo quality, but provides no published validation metrics, test set performance, baseline comparisons, or correlation analysis with actual social engagement data. The product description asserts virality prediction capability, yet the architectural analysis reveals no evidence of training on real social media performance data or validation against ground truth engagement metrics. The scoring function appears to be a proprietary blend of these four dimensions, but the weighting, feature extraction, and prediction methodology are entirely undocumented.

Uploads images to Vercel-hosted infrastructure for server-side processing, but provides no documented data retention policy, deletion mechanism, or privacy guarantees beyond a vague 'Private & secure' claim. The system does not specify whether uploaded photos are stored permanently, cached for reanalysis, deleted immediately after processing, or retained for model training. No mention of GDPR compliance, data export capabilities, or user deletion rights. The privacy model is entirely opaque, creating significant risk for users uploading personal photos (especially sensitive profile pictures or dating app images).

Restricts all analysis operations to web UI interactions — no REST API, GraphQL endpoint, or SDK is mentioned or available. Users cannot programmatically upload images, retrieve results, or integrate Hotcheck into automated workflows. This architectural choice prevents batch processing, integration with photo management tools, or CI/CD pipelines. Each analysis requires manual web UI interaction, making the tool unsuitable for creators managing large photo libraries or teams needing scalable analysis infrastructure.

Returns a single blended 'hotness rating' without breaking down which visual dimensions (face, body, style, photo quality) contributed to the score. The system claims to evaluate four distinct dimensions but provides no per-dimension scores, feature importance weights, or visual explanations (e.g., 'face attractiveness: 7/10, body attractiveness: 6/10'). Users receive only a composite number and mode-dependent text insights, with no ability to understand which aspects drove the rating or how to improve specific dimensions. This lack of explainability makes the tool a black-box engagement vanity metric rather than actionable feedback.