Skip to content
SiteList

Erase.tools Review: strong cleanup tool, thin self-serve help (80/100)

Erase.tools scores 80/100, with standout usability and accessibility around a clear browser-based cleanup workflow. Performance is a clear constraint, with homepage LCP at 3.0 s at p75 and CLS at 0.14.

Reviewed by SiteList Engine · 13 dimensions · published Reviewed on October 27, 2026

Is this your site?

Claim it

Quick facts

Erase.tools is a freemium, browser-based AI-assisted photo and video cleanup product reviewed across 35 crawled pages on 2026-10-19.

Fact Value
Domain erase.tools
Category AI-assisted photo and video cleanup tools
Pricing Freemium
Price range ?–? USD; no numeric plan price supplied
Pages crawled 35
Crawl date 2026-10-19
Evidence
Pages crawled
35
Crawl date
2026-10-19
Pricing model
freemium

Executive summary

Erase.tools is a strong overall product at 80/100, with its clearest advantages in usability, accessibility, technical SEO, and risk and stability. Usability scores 96/100: all six desktop and mobile persona runs reached their goal. Accessibility scores 98/100, while technical SEO scores 93.6/100 across the 35-page crawl.

The product is easy to classify and try. Audience and messaging scores 76.6/100, first impressions and positioning score 74.3/100, and design execution scores 82/100. Writing quality is 83.9/100, decision-support surfaces are 88/100, review-content integrity is 86/100, and editorial QA is 76.3/100.

The two clearest weaknesses are performance and self-serve help. Performance scores 61.5/100: homepage LCP is 3.0 s at p75 and CLS is 0.14, although INP is 111 ms.

Evidence
Overall score
80/100
Usability
96/100
Accessibility
98/100
Performance
61.5/100; LCP 3.0 s p75; CLS 0.14; INP 111 ms
Docs & self-serve help
36.5/100; no public help-center or documentation subtree found

01 · First impressions & positioning — box-and-rebuild promise, 500 MB support, no named buyer

Erase.tools is easy to classify because its hero states both the job and the mechanism: “Remove text, captions, subtitles from video,” then “Box the text and erase.tools rebuilds the background behind it.” The supplied homepage evidence adds before/after media, 500 MB support, MP4/MOV/WebM compatibility, a 10-credit free trial, and full-resolution output. That makes the promise concrete rather than generic. The gap is audience specificity: the evidence names tasks, but no role, industry, or company-size segment. Proof is also mainly demonstrative; no attributable customer outcome or review count is supplied. Keep the task-led hero, then add one audience cue and one attributable example with file type, cleanup difficulty, and result.

Evidence
Homepage hero claim
Remove text, captions, subtitles from video; box the text and rebuild the background
Supported upload limit
500 MB
Free trial
10 video credits

02 · Audience & messaging — clear first task, paid cost still unresolved

Erase.tools answers the first-use questions quickly: what it does, how the interaction works, and where to begin. The homepage explains the box-and-rebuild workflow, shows before/after proof, and places “Get started for free” beside MP4/MOV/WebM, 500 MB, and trial details. The FAQ then covers moving watermarks, burned-in subtitles, batches, exports, and API access. Decision friction remains around paid cost: Pro is described as 600 to 900 video credits per month based on billing term, but no currency example is supplied. The highest-value messaging improvement is a worked plan example—term, price, included credits, and a representative cleanup—paired with a named primary user role.

Evidence
Pro allowance
600 to 900 video credits each month, based on billing term
Homepage start CTA
Get started for free
Batch capacity
Up to 50 files at once

03 · Usability — six goal-complete persona runs, 33 homepage links

Erase.tools has a direct first-use path: the homepage identifies the task within ten seconds, keeps a free-start action visible on desktop and mobile, and sends a visitor to /caption-remover in one click. All six sampled desktop and mobile persona runs reached their goal; the skeptical purchase path took one extra step through the pricing cards. Labels such as “Remove text from image,” “Photo watermark remover,” and “Caption & subtitle remover” match ordinary visitor goals. The main friction is breadth: the homepage exposes 33 outgoing links, while the mobile lab reports 4.6 s LCP and 4.9 s TTI. Keep one dominant hero action, group the catalog by job, and reduce mobile entry cost.

Evidence
Persona runs reaching goal
6 of 6
Homepage outgoing links
33
Mobile homepage lab timing
4.6 s LCP; 4.9 s TTI

04 · Accessibility — 97 PSI homepage score, with contrast failures across three templates

Erase.tools shows a strong automated accessibility baseline, but color contrast is the confirmed improvement area. Six sampled rendered-DOM records include header, nav, and main landmarks; the homepage also includes footer. They report one h1 with orderly h2/h3 progression, zero images missing an alt decision, zero unlabeled inputs, and zero positive tabindex values. The supplied stylesheet contains 9 :focus rules, 3 :focus-visible rules, and 1 reduced-motion rule. Against that baseline, PageSpeed Insights reports accessibility scores of 97 for the homepage and 96 for /ai and /mcp, with the color-contrast audit failed on all three. Retrieve the affected nodes and ratios, fix shared tokens or styles, then rerun the audit.

Evidence
PageSpeed accessibility
97 homepage; 96 on /ai and /mcp
Sampled DOM label findings
0 inputs without a label or ARIA name
Focus-rule inventory
9 :focus; 3 :focus-visible; 1 prefers-reduced-motion

05 · Design execution — 36×36 px mobile header control, stable 0 CLS

Erase.tools presents a coherent visual system in the supplied captures: the desktop hero moves from headline to supporting copy, blue “Get started for free” CTA, and before/after demonstration, while “Explore all tools” remains secondary. The mobile layout stays in one column, with a 390 px scroll width and 16 px body text. The homepage lab records CLS of 0 on both mobile and desktop. The clearest mechanical defect is the mobile header control: it measures 36×36 px at the 390 px layout, and the same report counts 40 small tap targets. Expand the interactive box to at least 44×44 px without enlarging the icon. Treat the 15 font-size and 9 radius values as drift risk, not proof of visible inconsistency.

Evidence
Mobile header control
36×36 px at 390 px layout
Homepage mobile tap targets
40 small targets
Homepage lab CLS
0 on mobile and desktop

06 · Performance — 3.0 s field LCP, 2.19 s TTFB, 1,833 KiB homepage

Erase.tools is responsive once loaded, but its main entry path misses key mobile performance targets. Real-user PSI data records homepage p75 LCP at 3.0 s and CLS at 0.14, while INP is a strong 111 ms. Field TTFB is 2.19 s, and the mobile homepage transfers 1,833 KiB across 64 requests. Lab LCP reaches 4.6 s on mobile versus 0.7 s on desktop; /ai reaches 5.2 s in the mobile lab. The evidence estimates 175 KiB from GTM and 97 KiB of unused JavaScript savings. Trace the hero request and server path first, then defer non-critical work and reduce the initial payload.

Evidence
Homepage field p75
3.0 s LCP; 0.14 CLS; 111 ms INP
Homepage field TTFB
2.19 s
Mobile homepage transfer
1,833 KiB across 64 requests

07 · Writing quality — concrete limits, 25.9-word average on the longest page

Erase.tools writes most clearly when it pairs the job with its boundary. The homepage explains the rebuild mechanism, while product pages use concrete cases such as photobombers, date stamps, license plates, captions, and document shadows. “Where it will let you down” and “What blur does not do” make limitations visible instead of hiding them. The weakness is catalog repetition: sampled pages reuse the same sequence of sign-in prompt, use cases, workflow, comparison, limitations, questions, and related tools. On /object-remover, rendered text measures 2,384 words across 92 sentences, with a 25.9-word average. Keep the scaffold, split long sentences, and require one page-specific rule or worked example before shared sections.

Evidence
Object-remover rendered text
2,384 words; 92 sentences; 25.9-word average sentence
Repeated catalog sequence
Sign in, use cases, how it works, comparison, limitations, questions, related tools
Homepage missing-alt findings
2 images without alt text

08 · Decision-support surfaces — 90-second limit and 500 MB cap drive the choice

The CapCut comparison gives visitors a usable decision rule rather than a one-sided feature list. Its “Your situation” table routes clips longer than 90 seconds, footage that must stay on the machine, and work needing cutting or grading afterward to CapCut; tracked watermarks, repeated platform stamps, and exports that preserve untouched pixels favor erase.tools. The matrix also surfaces browser versus desktop, server upload versus local processing, and a 500 MB maximum file size. Fourteen feature rows make the first scan denser than necessary. Keep privacy, clip limits, tracking, export fidelity, and editing workflow at the top, and add a dated “How we compared” note beside the table; the page currently provides sources and a 1 September 2026 publication date.

Evidence
Comparison feature rows
14
Erase.tools maximum clip length
90 seconds
Erase.tools maximum file size
500 MB

09 · Review-content integrity — balanced tradeoffs, methodology disclosure still needed

The supplied CapCut comparison has a credible evidence posture because it recommends the alternative when the workflow calls for it. It names longer than 90-second clips, local-only footage, real-object removal, and continued editing as reasons to choose CapCut, while assigning moving-watermark tracking and untouched-pixel export to erase.tools. The page links sources, uses ItemList schema for the two options, and the sampled evidence includes no affiliate links, Review markup, or AggregateRating markup. The missing piece is a compact methodology statement: readers are not told which criteria were selected, which limits are published facts, or whether hands-on testing occurred. Add that disclosure and dated what-changed notes when comparison facts move.

Evidence
Comparison publication date
1 September 2026
Structured comparison type
ItemList describing two software options
Sampled affiliate links
None supplied in the crawl evidence

10 · Risk & stability — indexable 35-page catalog, but a two-hop canonical redirect

The current crawl shows no supplied traffic-loss symptom and no broad indexability failure. Public inventory URLs return 200 with self-canonicals, robots.txt leaves the public site crawlable, the nonexistent probe returns 404, and raw and rendered word counts match across the crawl. The main operational risk is the host path: http://erase.tools/ redirects to https://erase.tools/, then to https://www.erase.tools/, costing two 308 hops. A second exposure is concentration: 33 of the 35 inventoried pages are tool pages around one media-cleanup theme. Make the canonical redirect one hop, then add genuinely useful creator, editor, or marketing workflow pages with distinct utility rather than near-duplicate tool pages.

Evidence
Crawled pages
35
HTTP canonical redirect chain
2 permanent redirects: HTTP root → HTTPS non-www → HTTPS www
Tool-page inventory
33 of 35 pages

11 · Editorial QA of content — repeated anchors at 32, with claims needing a source register

Erase.tools clears the basic human-editor bar: its copy names objects, scenes, file types, privacy boundaries, and failure cases, while the FAQ gives operational limits such as 50-file batches. The QA gap is systemic repetition. Internal anchor data records counts of 32 for repeated strings such as “object removerErase anything from a photo” and “remove text from imageCaptions and date stamps.” Capability claims also need a maintained source-of-truth gate: the sample states that a 4K clip comes back 4K and that users can drop up to 50 files, but supplied evidence includes no source-verification result for those claims. Vary contextual anchors, add a source register, and fix repeated missing-alt findings.

Evidence
Repeated anchor count
32 for sampled catalog anchor strings
Output claim
A 4K clip comes back 4K
Batch claim
Drop up to 50 files at once

12 · Docs & self-serve help — FAQ and API guidance on the homepage

Erase.tools surfaces useful answers in its homepage FAQ. It covers burned-in subtitles, video text removal, batch jobs, credits, exports, and integrations, and exposes FAQPage markup. The machine-readable /llms.txt index is a useful start, but it conflicts with the homepage: it says there is no public API, while the FAQ describes an API and JS/Python SDKs. Reconcile the API statement and add visible update dates.

Evidence
Homepage help format
FAQPage with answers on subtitles, batches, credits, exports, and integrations
API statement conflict
Homepage describes API and JS/Python SDKs; /llms.txt says no public API

13 · Technical SEO — 35-URL sitemap, self-canonicals, mobile LCP still needs work

Erase.tools has a sound crawl and rendering base. The verified sitemap lists 35 URLs; robots.txt allows the public site while limiting /api/ and /billing/; sampled public pages return 200 with self-referencing canonicals; and /sign-in plus /sign-up are noindex, follow and outside the sitemap. The guaranteed nonexistent URL returns 404, and raw and rendered word counts match, so the crawl found no evidence of a JavaScript-only rendering dependency. The clearest technical fix is the two-hop HTTP-to-www redirect. Performance is the secondary SEO opportunity: mobile field LCP is 3,013 ms with CLS 0.14 on the homepage, while INP is 111 ms. Keep the sitemap and robots directives aligned as new pages ship.

Evidence
Verified sitemap
35 URLs
Homepage mobile field data
3,013 ms LCP; 0.14 CLS; 111 ms INP
Account-page indexing
/sign-in and /sign-up are noindex, follow and excluded from sitemap

Verdict — 80/100: strong cleanup tool, thin self-serve help

Erase.tools is a strong cleanup tool for creators, editors, marketers, and other media owners who need to remove unwanted text, captions, watermarks, objects, or backgrounds from existing files. Its best evidence is practical: usability scores 96/100, accessibility scores 98/100, and the product facts describe browser-based cleanup with batch support and assistant integrations.

One fix should come first. Performance scores 61.5/100, with homepage LCP at 3.0 s p75 and CLS at 0.14. Improving performance would make an already easy-to-try product easier to adopt and sustain.

The fit is strongest for people who want direct file cleanup in a browser and a free trial before committing to credit-based Pro plans.

Evidence
Overall score
80/100
Usability
96/100
Accessibility
98/100
Docs & self-serve help
36.5/100
Performance
61.5/100; LCP 3.0 s p75; CLS 0.14

Methodology & data notes

This 13-dimension review uses the 35-page crawl dated 2026-10-19, supplied product facts, the public dimension score table, and the recorded enrichment state. The supplied public score table covers dimensions 01, 02, 03, 04, 05, 07, 09, 12, 13, 17, 19, 23, and 25.

Dimensions not represented in that table are excluded from this article frame because no public score summaries were supplied for them; they are not treated as failures. Search Console enrichment is pending: the supplied site profile records both gsc_connected and gsc_access as false.

This is a 13-dimension review, and its claims stay within the supplied crawl and product evidence. Read the full SiteList methodology for the scoring and data notes.

Evidence
Review scope
13 dimensions
Crawl scope
35 pages on 2026-10-19
GSC enrichment
gsc_connected=false; gsc_access=false

Questions buyers actually ask

Does the output lose quality?

No. The product says it rebuilds only the selected pixels and leaves the rest of the frame untouched, including resolution, bitrate, and container.

Can it follow a moving watermark?

Yes. Tracking follows the mark across the timeline, so a moving watermark needs to be marked once. Long clips with heavy motion still merit a quick review before export.

What about burned-in subtitles?

You can draw one box over the caption band and clear it across the clip. The supplied product facts say results are usually indistinguishable from a clean master on flat areas.

Can I run a batch?

Yes. You can drop up to 50 files at once and apply the same selection region across them, which suits reposts and localized cuts of one video.

Is there an API?

Yes. The supplied FAQ describes one POST with a file and region, followed by a signed URL, with JavaScript and Python SDKs and per-second pricing.

Can I remove text from a video for free?

A free account includes one 10-credit video trial. Pro includes 600 to 900 video credits each month, based on the billing term.

How this review was made

SiteList reviewed erase.tools on October 27, 2026 — pages, screenshots, performance runs, structured data and public records — then scored it across all 13 public dimensions. Every claim above is sourced from what we collected; nothing is hand-tuned and the score is never for sale.

Pending enrichment (data we could not fetch this run): SERP competitor lookup was unavailable because no SERP endpoint was configured., The supplied crawl is partial (35 of 40 pages), so cross-page messaging consistency is not fully assessable., The available SERP endpoint was not configured, so audience phrasing and competitor language could not be validated externally., Only the supplied homepage evidence was used for the detailed question audit; the 35-page crawl count does not itself expose the content of every page., Live Playwright execution was not provided; clickability and form validation were assessed from crawl inventory, link graph, copy, metrics, and screenshots., Detailed contrast nodes, color pairs, and ratios for the failed PSI audits., Live keyboard/tab-order verification and screen-reader announcement testing., Manual verification of modal focus trapping, dynamic error announcements, and media caption accuracy.…

Read the full methodology

80/100EraseJump to review