Reactor Review: capable API, costly performance (67/100)

Field	Value
Domain	reactor.inc
Category	AI developer platform / real-time video generation API
Pricing	Usage-based; no tier range supplied in the evidence
Pages crawled	32 of 40
Crawl date	2026-09-01

Reactor is a developer platform for real-time video generation with a 67/100 overall score. Its strongest area is Docs & self-serve help at 93/100: the documentation is comprehensive, technical, and tailored to real-time AI video developers, with deep model API references and clear JavaScript and Python examples.

The central weakness is performance at 32/100. Payloads exceed 5MB, field LCP reaches 6.7 seconds, JavaScript execution takes 4.6 seconds, and render-blocking resources compound the delay. Usability scores 78/100, but content discovery on the marketing site remains broken in the supplied walkthrough. Audience and messaging score 62/100, reflecting a clear developer audience but a weak 10-second category test. Technical SEO and writing quality each score 74/100, while risk and stability scores 72/100. The next gains come from reducing page weight and repairing discovery and rendering paths.

Reactor presents a clear developer platform for real-time video generation, but its name creates discovery friction. The hero targets AI engineers and full-stack developers with the phrase "The developer platform for real-Time world models" and backs it with three concrete claims: <1s round-trip latency, zero infrastructure to manage, and 1,000s of GPUs on demand. Those claims have no source links or methodology notes, and the supplied crawl evidence found no testimonials or case studies on the sampled surfaces. Add "API" or "Video Generation" to the H1, navigation labels, and meta titles, then link each performance claim to supporting evidence.

Reactor speaks convincingly to AI engineers and full-stack developers, but broader application developers must translate the academic term "world models" before they understand the product category. JavaScript/React/Python SDK references, the reactor.sendCommand('set_prompt', ...) example, and a Build with AI agents link establish who can start. The supplied crawl did not show pricing tiers on the primary docs surface, while Billing is available through navigation. Lead with "Live Video Generation API", surface pricing within one click of the docs overview, and retain "world models" as technical context.

Core developer onboarding is easy to reach, while content discovery has a hard failure. A simulated walkthrough reached /quickstart and /authentication in 2 clicks, and the billing page uses anchors for pricing mechanics, models, and buying credits. In contrast, the /blog inventory reports links: [] and buttons: [], leaving visitors with no navigable list of updates. Pricing also routes from the marketing site to docs.reactor.inc/resources/billing, creating a context switch. Populate the blog index and mirror a concise pricing path on the main site.

Reactor’s sampled docs templates provide header, navigation, main, and footer landmarks, and all 5 sampled images have an alt decision. The main barriers are structural: the homepage and careers footer each contain an email input with placeholder YOUR EMAIL but no label or ARIA name; the docs heading sequence begins h2 → h3 before reaching h1; and sampled pages report no skip link. The blog and status samples also lack a main landmark. Add accessible names, place one descriptive h1 before subordinate headings, add a keyboard-visible skip link, and wrap primary content in main.

Performance is the clearest constraint on Reactor’s first load. Field LCP is 6,713 ms and lab mobile LCP is 7,454 ms, against a 2,500 ms target. The mobile page weighs 5,749 KiB, while the /home desktop template reaches 12,033 KiB; JavaScript execution takes 4.6 seconds and render-blocking resources carry an estimated 3,170 ms saving. Compress images to AVIF/WebP, defer non-critical scripts, split route bundles, and remove the reported 693 KiB of unused JavaScript and 297 KiB of unused CSS.

Reactor’s strongest copy is concrete: the homepage pairs "Full-stack infrastructure for real-time video and world models" with code, while the careers page uses $59M raised and 2025 founded as trust signals. Billing also explains pay-as-you-go pricing in operational terms. Editorial discipline breaks on the blog and status surfaces: the blog has no headings and only the text "Blog — Reactor", while the status page lacks an h1 and meta description. Three routes—/auth, /legal/terms, and /legal/privacy—share one meta description. Restore substantive introductions and write page-specific metadata.

The technical foundation is functional, but content delivery depends too heavily on client-side rendering and redirect-heavy URL hygiene. The blog falls from 202 raw words to 3 rendered words in the crawl snapshot, and /robotics is 96.1% JavaScript-dependent. The sitemap includes 307 redirects for blog, model, and API paths, while the HTTP root takes two hops before reaching the preferred URL. Move primary content to server rendering or prerendering, emit final canonical URLs in the sitemap, and consolidate the HTTP redirect chain.

Reactor’s documentation is its most complete surface, with model API references that pair event definitions and SDK setup with 140 to 260+ code blocks per page. The onboarding path includes npx create-reactor-app, JavaScript and Python package installs, billing guidance, and an 8KB /llms.txt with model listings and pricing matrices. The main limitation is fragmentation: platform guides and billing live on docs.reactor.inc, while model API references live under reactor.inc/models/*/api. Add unified search and navigation, then publish a central changelog.

Technical SEO is broadly indexable, but rendering and canonical URL handling create avoidable delays. The blog snapshot contains 202 raw words but only 3 rendered words, and /robotics has a 96.1% JavaScript-only share. The sitemap lists 307 redirects, including blog and model paths, instead of final canonicals. In addition, http://reactor.inc/ resolves through two hops, and /home is marked noindex while remaining in the sitemap. Server-render primary content, emit final URLs only, consolidate redirects, and remove /home unless it is intentionally indexable.

Reactor is best suited to software developers and AI engineers building generative media applications who need a real-time video generation API. The product case is clearest in its technical proof points, SDK examples, and documentation, which earns 93/100 for Docs & self-serve help.

The priority is to reduce oversized payloads and improve field LCP, then repair the marketing site's content discovery path, rendering dependencies, and sitemap hygiene. Those changes would make the platform easier to discover and faster to evaluate.

This 13-dimension review uses the supplied public dimension score table and crawl evidence for reactor.inc. The crawl covered 38 pages on 2026-09-01. Google Search Console access was unavailable.

Dimensions 05 (Design execution) was excluded from the supplied scope; 12 (Decision-support surfaces) and 13 (Review-content integrity) were not applicable; and 19 (Editorial QA of content) was missing from the supplied scope. The review does not add sections for excluded dimensions.

Read How SiteList scores for the review framework.