Ranking method
How we rank tools for your project.
Our tiers recommend tools for specific tasks based on documentation. Availability, commercial licensing and content policy are listed separately: a high rank does not establish permission for your game’s content.
What the tiers mean
How we choose a tier
We review primary documentation, model cards, licenses, provider policies and availability notices. For each workflow, we consider documented controls, repeatability claims, editing options, integration requirements, learning resources and costs. The tier reflects our editorial judgment of that fit. It is not a vendor rating or a measured comparison of output quality.
Model tiers apply to permitted non-explicit development tasks. An A-tier concept tool may prohibit the explicit imagery your finished game needs. N can mean an endpoint has retired or a license permits only research; it does not establish that the model produces poor images.
What the evidence covers
The catalog has been reviewed against documentation. We have not run model-generation benchmarks, engine comparison benchmarks or commercial asset-production tests for these ranks. Numeric scores remain unassigned. Website checks cover navigation, data and page behavior; they provide no hands-on evidence of model output or engine performance.
A practical test report would identify exact model or tool versions, settings, hardware or hosted endpoints, and a fixed non-explicit test set of owned or consented material. It would measure the share of outputs accepted for the stated task, correction time, quality after integration and full production cost. Proposed numeric weights are workflow fit 30%, result quality 20%, repeatability 20%, learning and documentation 15%, and total cost 15%. Those weights have not been applied to this catalog.
Check permissions separately
Restricted means a documented policy limits the relevant content. Conditional means permissions depend on the exact license, intended use or additional agreements. Unverified means the available sources do not establish permission for the proposed project. These labels provide no general legal clearance. Running a tool locally does not remove its applicable license or acceptable-use conditions.
Commercial use, fine-tuning, redistribution of model weights and use of generated outputs can each have different terms. Territorial restrictions can affect outputs as well as access. Check the exact version’s terms and the intended destination of its outputs before committing to a pipeline.
Coverage and other rankings
The directory covers major documented image, video and 3D families, meaningful differences between services, weights and versions, representative community baselines and useful legacy references. Coverage is selective: we cannot list every checkpoint, LoRA, hosted wrapper or private beta. Each entry explains which variants it groups. Unavailable products remain listed so their status is clear.
Guide reading priority is S for common foundations, A for focused production tasks, and B for advanced specialist work. Resource ranks help you choose a first, supplementary or specialist reference for a subject. Publishing ranks point to routes worth investigating for the stated use. Each ranking answers a different question; combining them into one overall score would hide those distinctions.
Updates and corrections
The proposed maintenance schedule calls for monthly reviews of provider policies and availability, plus another check before an affected release. Tool recommendations need a quarterly review; tutorials need a review after relevant engine changes. Each entry records the date and sources used. This schedule describes planned editorial work, not an automatic monitor or a record of completed reviews.
Updates should preserve historical version distinctions. A changed endpoint or license needs a supporting source, a revised recommendation and a check of linked guides. Affiliate or sponsored placement must be disclosed and must not determine rank.