mutationdb.com

What It Is

FlavorLab is an AI Sensory Prediction tool for food formulation. Users can paste a recipe, upload a file, or submit a URL, and the system predicts a full sensory profile. Through FlavorLab Assistant, users can also explore in conversation how changes in ingredient ratios may affect sensory outcomes. Judging from the page flow, it sends content to Claude AI, extracts structured ingredients, then queries a sensory database, researches unknown ingredients, and runs Lasso models.

Core Capabilities and Outputs

The product is not a general-purpose chatbot; its core focus is vertical prediction around “recipe—ingredients—sensory profile.” Example prompts include increasing sugar to 25%, removing butter, adding 10% olive oil, or asking why something tastes bitter. In terms of output, the page shows fields such as Ingredients & Volume Fractions, Model Confidence, Predicted Scores, Score/100, predicted ± RMSE ranges, and Ingredient Details including weight, source confidence, and evidence. These fields help R&D teams judge whether predictions are trustworthy and make the results more explainable than plain text suggestions.

Pricing, Chinese Support, and Integration

The captured text does not disclose any free quota, trial, subscription pricing, or payment methods. There is also no visible information about an API, enterprise integration, batch processing, or team collaboration. Chinese support is likewise not clearly stated. While Claude theoretically supports multiple languages, the page does not promise Chinese recipe parsing or Chinese output quality, so its suitability for Chinese users cannot be directly assessed.

Pros, Cons, and Best-Fit Users

Its main advantage is its focused use case: it is suitable for food R&D teams, chefs, flavor researchers, and formulation developers who want to quickly compare the sensory impact of different recipe adjustments. The inclusion of confidence scores, RMSE, and evidence also makes it useful for preliminary screening. The downside is the lack of public information, especially around pricing, privacy policy, whether data is used for training, database coverage, and model validation methods. Sensory prediction itself also cannot replace real tasting panels, so it is better suited to early-stage exploration than final conclusions.

Access from China and Alternatives

Access from mainland China is unknown. The page indicates reliance on Claude AI, so real-world usability may be affected by network conditions and upstream model availability. Payment methods are also not disclosed. If stable access is not possible, alternatives include using general-purpose large language models together with a self-built recipe database, food R&D software, or a sensory evaluation workflow, though these options usually require more manual modeling and validation.