SubtitleGenerator scores 82/100 for a clear browser-based subtitle workflow that makes uncertain words easy to find and fix. Its main weakness is mobile performance: lab LCP ranges from 4.1s to 6.2s, and the homepage weighs 5.7MB.
Reviewed by SiteList Engine · 12 of 13 dimensions · published Reviewed on November 7, 2026
Is this your site?
Claim itSubtitleGenerator is a browser-based AI subtitle generator for creating, fixing, styling, translating and exporting video subtitles.
| Fact | Detail |
|---|---|
| Domain | subtitlegenerator.app |
| Category | AI subtitle generation, correction, styling, translation and video export |
| Pricing | Free $0; Pro $9/month; Max $15/month; PAYG from $3 |
| Pages crawled | 38 |
| Crawl date | 2026-10-25 |
SubtitleGenerator is a strong browser-based subtitle product with clear positioning, high usability and solid supporting foundations. Its clearest strength is the correction workflow: it generates subtitles, then makes uncertain words easier to find and fix.
The strongest public scores include risk & stability at 93/100, technical SEO at 90/100, usability at 89.2/100 and accessibility at 88/100. The main opportunities are practical. Performance scores 75.3/100, with lab mobile LCP ranging from 4.1s to 6.2s and a 5.7MB homepage. Docs & self-serve help scores 68/100, while decision-support surfaces score 73/100. Writing quality remains strong at 79.3/100, but the editorial program still needs further refinement before senior-editor sign-off.
The result is a product that explains its core job well and gives buyers useful evidence about starting, editing and exporting subtitles, while leaving clear work to do on first-load speed and the depth of self-serve guidance.
SubtitleGenerator establishes a clear category and a specific mechanism: it generates subtitles, flags uncertain words, and lets people fix them quickly. The hero says "Free AI subtitle generator that's actually fast to fix" and follows with "Generate subtitles with AI, then fix the words it got wrong in seconds." That is more useful than a generic AI-caption claim because the failure mode and remedy are visible. The audience is still inferred from examples for podcasts, tutorials, gaming, courses and short-form platforms rather than named above the fold. Keep the correction territory, but add a creator-and-publisher qualifier and one attributable workflow result beside it.
A prospect can understand what the product does, how to begin, what it costs and what happens to the video without leaving the public site. The homepage connects upload, transcription, flagged-word review, styling and export, while the pricing page states Free at $0, Pro at $9/month and Max at $15/month with limits for minutes, video length, file size, watermark behavior and export quality. The message is clear, but the persona remains assembled from task labels and platform contexts such as TikTok, YouTube, CapCut, Premiere and DaVinci. Add one concise creator or publisher qualifier, then support the correction claim with attributable customer outcomes rather than more product-authored proof.
The core self-serve path is easy to follow: the upload surface appears near the top, a 20-second sample offers a low-commitment start, pricing exposes a comparison table, and the tools taxonomy uses direct task labels such as Generate, Review, Translate, Style, Learn and Publish. The main friction is support discovery. All three contact walkthroughs began at the homepage, required a long scroll, and ended at the footer's Contact us link before reaching the short labeled form. Put Contact or Support in primary or utility navigation, and choose one primary hero label so Start free, Upload and the sample clip have a clearer hierarchy.
SubtitleGenerator has a solid semantic baseline in the sampled public pages: header, nav, main and footer landmarks are present, and all 32 sampled homepage images have non-empty alt attributes. The remaining issues are concrete. PSI reports a failed color-contrast audit on the homepage, pricing page and subtitle-design guide, plus a homepage label-content-name-mismatch. The 390px layout facts also report 24 small tap targets, including a 36×36px header menu button and 24×24px footer icon buttons. Add a focus-visible skip link, identify the failing nodes and enlarge those hit areas to at least 44×44px before re-running the audits.
The visual system is coherent across the supplied homepage, tools and pricing captures: an oversized hero leads into the upload surface, cream, black and yellow carry emphasis, and the tools cards stack into a single column at 390px without measured horizontal overflow. The score is held back by mechanical details rather than a confused direction. The mobile facts report 24 small targets, including the 36×36px header menu and 24×24px footer icon controls; pricing also has a heading-order failure and the homepage has an accessible-name mismatch. Preserve the existing hierarchy while enlarging hit areas, repairing the semantic outline and consolidating the actually used visual tokens from the broad extracted inventory.
Mobile loading is the clearest experience constraint in the lab sample. LCP measured 5,251 ms on the homepage, 4,070 ms on pricing, 4,801 ms on the subtitle-design guide and 6,226 ms on the closed-caption article, while desktop LCP stayed between 801 ms and 1,001 ms. The homepage also sent 5,703 KB across 114 requests, with 41 blocking requests, at least 15 third-party domains and 154 font-face entries across 15 families. Start with the verified LCP candidate, then defer below-the-fold work and non-essential integrations. CLS was 0 throughout, and no CrUX field data was supplied.
The strongest copy names the product job and its failure mode plainly: "Generate subtitles with AI, then fix the words it got wrong in seconds." The main editorial weakness is structural precision. The how-to page promises four steps, then continues through H2s numbered 1 through 10, forcing readers to reinterpret the article's shape. Make four top-level workflow steps and move supporting material underneath, or rename the promise to match the ten-part structure. The homepage also averages 26.4 words across 84 sentences, so splitting long feature sentences would improve scanability. Keep editor states useful for people, but consolidate their repeated search-facing metadata around the homepage.
The editorial comparison pages guide decisions well: closed captions are positioned for accessibility and easy editing, open captions for muted autoplay and feed-first platforms, and the companion comparison adds audience, dialogue, sound and language context. Pricing is less decisive. The $0, $9 and $15 plans, annual prices and PAYG from $3 are transparent, but the 17-row grid does not say who should choose each tier or name the trade-off for moving up. Add concise fit labels—Free for trying the editor, Pro for regular watermark-free work, Max for longer or higher-volume projects, PAYG for overflow. On mobile, replace or clearly expose the clipped comparison table; the Max column is only partly visible.
The scan found no confirmed active traffic-suppression defect in the supplied evidence. The public inventory records 200 responses, robots.txt allows the public root, the verified sitemap contains 38 URLs, the nonexistent-URL probe returns a real 404, and HTTP and www consolidate to HTTPS in one hop. Raw and rendered content is effectively identical across the 38-page sample. The material risk is mobile speed: lab LCP is 5,251 ms on the homepage, 4,070 ms on pricing and 4,801 ms on the subtitle-design guide. Treat that as a conversion and future traffic risk, not proof of loss; connect GSC and analytics before attributing a change.
The corpus is specific and useful, but the flagship how-to should not pass senior-editor sign-off unchanged. Its heading says "How to add captions to a video in four steps" while the outline runs through ten numbered H2s. Align the contract before publication by making four sections the actual workflow with supporting guidance nested beneath them, or change the opening promise. A second QA gate concerns editor and sample query states: the homepage title and meta description are repeated across eight pages, and sampled titles reach 69 and 75 characters. Consolidate duplicate metadata and trim only the longest titles. The supplied evidence confirms neither a schema mismatch nor a dead citation, so neither is presented as a defect.
The public guides are practical but narrow for a browser-based editor. The design guide is 808 words and the guidelines guide is 1,481 words; together they cover presets, size, weight, outline, contrast, safe area, line length, publisher rules and accessibility checks. Both sampled guide pages have breadcrumbs and side navigation, but neither has a search box, and the crawl does not surface a dedicated getting-started or task-help index. Keep the recent material and the 4,939-byte /llms.txt foundation, then add a Guides or Help hub organized around editing, styling, translation, export and troubleshooting. A scoped search field would make those answers easier to reach.
SubtitleGenerator has a strong crawl and rendering foundation. robots.txt declares the sitemap and limits private paths, the verified sitemap contains 38 URLs, the nonexistent-URL probe returns a real 404, and the public inventory records 200 responses with coherent canonicals. The raw-versus-rendered sample covers 38 pages and reports zero JavaScript-dependent pages; /tools also exposes WebPage and BreadcrumbList JSON-LD, and /llms.txt returns 200. The remaining issues are experience and URL hygiene: pricing has 473px scroll width at a 390px viewport, and /tools/ is served as a second 200 URL instead of redirecting to the chosen trailing-slash form. Fix those after the mobile LCP path.
SubtitleGenerator earns its 82/100 through a clear category, a credible correction differentiator and a highly usable front door. The public site explains what the product does, how to start and what happens to the video; the product facts also make the editing and export boundaries clear.
Three fixes would make the decision easier: reduce mobile first-load time, address the 4.1s–6.2s LCP range and 5.7MB homepage weight, and deepen the self-serve help surface, which scores 68/100. Decision-support surfaces score 73/100, so comparison guidance also has room to become more consistently useful.
This is a good fit for video creators, publishers, educators and social-video teams that need editable, styled subtitles in a browser. It is strongest for buyers who value quick correction and clear editing control.
This is a 13-dimension review based on a 38-page crawl completed on 2026-10-25, public product facts from JSON-LD and page text, simulated cognitive walkthroughs, sampled DOM and lab results, and the supplied editorial corpus. Scores and article claims use the public dimension table and its one-line findings.
Dimension 13, Review-content integrity, was excluded as not applicable. Google Search Console enrichment is pending: it is not connected and access is unavailable. No traffic symptom was supplied, so the risk & stability result is a vulnerability scan rather than a historical-loss diagnosis.
The Free plan has stated limits for subtitle minutes, video length, file size, watermark behavior and export quality. Check the pricing page for current limits.
No. Only the audio is sent for transcription and deleted afterwards.
SubtitleGenerator is described as supporting subtitle fixing, but the supplied evidence does not specify the available editing controls.