Test record & evaluation basis

Erase.bg · Background removal

Hair quality is good, but the HD signup flow failed to deliver, promotions interrupt the workflow, and prior images remain visible in a shared browser.

Test date 2026-10-15Free tier + creditsNative output attachedMethod 2.3.1
1 · Date, environment & tier2026-10-15 · Guest flow

US-West (California, AS2914)

2 · Fixture & native exportCurly-haired portrait (anchor_bgremove_v1)

1536 × 1536 · PNG · Open native file ↗

3 · Price checks vs. test scoreOfficial pricing checked 2026-10-20

Subsequent pricing and service checks are shown separately without altering historical scores.

4 · What one test cannot proveScoped to this fixture & path

One run does not establish all complex scenes, untested paid maxima or long-term uptime.

Plans & historical pricing

Current purchase cards on the Chinese pricing page; dollar pricing · 2026-10-20

Plan · source date
Monthly / one-time
Annual bill / period
Quota & limits
Signup allowance2026-10-20 ↗
Free trial
Trial
Trial 3 credits / 3 HD downloads after signup
10 credits2026-10-20 ↗
USD 6 · one-time payment
Usage / one-time
Usage / one-time 10 removals; 1 credit each Rules [1]
100 credits2026-10-20 ↗
USD 12 · one-time payment
Usage / one-time
Usage / one-time 100 removals; 1 credit each Rules [1]
1000 credits2026-10-20 ↗
USD 50 · one-time payment
Usage / one-time
Usage / one-time 1000 removals; 1 credit each Rules [1]
Enterprise2026-10-20 ↗
Contact sales
Quoted
Quoted Custom quote

5 plan / billing-period records, all listed; — means that period is not recorded.

[1] Valid for 1 year; non-refundable

The conflict persists on recheck: purchase cards offer only 10/100/1,000-credit one-time packs, while the FAQ claims unlimited Pro at $15/month or $5/month billed annually without a matching subscription purchase option. These subscriptions are not marked purchasable. Signup grants 3 HD downloads; each removal costs 1 credit; purchased credits expire after one year and are nonrefundable.

Verification has gaps; unknown prices and quotas are not inferred.

Cancellation, sources & verification record

Current packs are non-refundable. FAQ subscription cancellation text may describe legacy plans.

Official plans and usage terms ↗ · 2026-10-20 Capture 1 ↗

Current purchase cards and FAQ conflict recheck ↗ · 2026-10-20 Capture 1 ↗

Pricing recorded on the test date · 2026-10-15

Billed via PixelBin platform subscriptions or credit packs

Unified PixelBin platform credits and subscriptions

Historical context only; not current purchase guidance.

Delivery & access

Free delivery record
1536×1536 (75% dimension / 56.25% pixel area retained, no watermark)
Attached file dimensions
1536 × 1536 · PNG

Read directly from the attached file; this does not establish free download availability at that size.

Watermark
None
Free allowance
Guest tier exports at 75% dimension (1536×1536); registering a PixelBin account grants 3 credits on console.pixelbin.io
Account requirements
No login required for 1536×1536 export; HD button prompts Google / GitHub / Email login
Maximum specification record
1536×1536 guest export (75% of input dimension; 'Sign up & Download HD 2048×2048' button redirected to console.pixelbin.io without delivering the image)

May include advertised capabilities; this does not establish a tested paid tier.

Batch & API
Requires switching to PixelBin console for batch & API workflows
Manual refinement
Basic background color replacement on Erase.bg; no brush tool

Tested result & sample

Underlying segmentation model is outstanding (mad_band=0.0748, SAD(k)=27.9, #2 in raw hair edge accuracy), which makes its aggressive cross-product redirect dark pattern all the more unfortunate.

Edge error MAD
0.0739 · Lower is better
Global error SAD(k)
27.914 · Lower is better
Foreground retention
99.960% · Higher is better
Background residue
0.036% · Lower is better

These historical measurements describe alpha error at edges (MAD), total alpha error / 1000 (SAD), foreground retention and background residue. Low-resolution outputs are aligned to the reference under the original protocol; this does not imply a full-resolution export.

Fixture
anchor_bgremove_v1
Recorded environment
US-West (California, AS2914)
Network-observed server wait
8.3 s
End-to-end reference
30.64 s

Network wait includes round trips, queuing and polling error, not pure inference time. Local timings depend on the device; repetition counts are not standardized in legacy records.

Experience, rules & trust

Disclosure
1536×1536 downscaling and cross-product redirect to PixelBin are not disclosed before clicking
Interaction
Clicking the free download triggers two consecutive overlay modals (sign-up prompt + PostHog PixelBin survey popup), and the HD sign-up flow redirects across products.
Entity record
Shopsense Retail Technologies Ltd. (Fynd / PixelBin, Reliance-backed)
Origin
Mumbai, India
Domain history record
Launched 2021
Third-party tracking record
PostHog · Google Analytics
External information
Built by Indian retail-tech giant Fynd with an excellent segmentation model, but heavily penalized by our 60.0 hard cap due to unfulfilled HD sign-up redirect to PixelBin.
Processing
Cloud async polling API (api.erase.bg, 8.30s server time)
Retention statement
Note: Guest page retained and displayed prior session uploads from the same browser profile

A recorded statement does not verify deletion or all data-processing behavior.

Verifiable audit trail & evidence

Four SHA-256-verified stages link the benchmark fixture, same-run interface screenshots, native exported artifact and archived pricing sources.

Stage A · Benchmark fixture Curly-haired portrait

anchor_bgremove_v1 · Inspect fixture & file inspector →

Stage B · Same-run screenshots 3 flow screenshots archived

Captures entry rules, result view and export disclosure.

Stage C · Native exported file 1536 × 1536 · PNG

Open native exported file ↗

Stage D · Archived pricing sources 2 official source snapshots

Official plans and usage terms (snapshot ↗) · Current purchase cards and FAQ conflict recheck (snapshot ↗)

Screenshots, native output and official claims support different conclusions. Evidence grades are not product safety certifications.

Evaluation basis & full scores · Overall 60/100

Overall = honesty & reputation × 30% + capability × 70%. Component maxima follow this report’s rubric. List scores are rounded for display; original precision is retained here.

Scoring version 2.3.1 · Test fixture anchor_bgremove_v1

Weighted score 81.6 · Final score after review 60
Tested (2026-10-15 Pass 2): Clicking 'Sign up & Download HD (2048×2048)' and completing Google OAuth registration did not deliver the 2048×2048 download; instead, it redirected to a separate product (console.pixelbin.io general editor with 3 credits) while losing the processed Erase.bg image. Human review confirmed and applied the 60.0 hard cap.

Honesty & reputation · 63/100 · 30%

CriterionScoreRationaleBasis
Pricing transparency25 / 40 Guest download silently scales 2048×2048 input to 1536×1536 (75% dimension) without stating pixel dimensions on the button; clicking 'Sign up & Download HD (2048×2048)' and completing Google OAuth redirected to console.pixelbin.io instead of downloading the HD image. Evidence A →
Interaction restraint6 / 20 Clicking the free download triggers two consecutive overlay modals (sign-up prompt + PostHog PixelBin survey popup), and the HD sign-up flow redirects across products. Evidence A →
Background & reputation21 / 25 Shopsense Retail Technologies Ltd. (Fynd / PixelBin, backed by Reliance Industries, Mumbai, India). Evidence C →
Engineering & privacy11 / 15 Supports HTTPS, but in guest mode the page still displayed previous session upload history from the same browser profile. Evidence A →

Capability · 89.6/100 · 70%

CriterionScoreRationaleBasis
Task quality59.6 / 65 Transition band error mad_band=0.0739 · Global SAD(k)=27.914 · Foreground recall 99.96% / Background leak 0.036% Evidence A →
Deliverable specifications20 / 20 True Alpha channel (8/8); Max deliverable spec: 1536×1536 guest export (75% of input dimension; 'Sign up & Download HD 2048×2048' button redirected to console.pixelbin.io without delivering the image) Evidence A →
Processing speed10 / 15 Network-observed server wait: 8.3s (API mode: async; network upload 15.61s / end-to-end 30.64s shown for reference only, never penalized) Evidence A →

Legacy scores may include advertised maximum capabilities. Read tested output separately from paid capabilities. Records without matching tiers and scope are not ranked together.

Revisions & corrections

Prepare a correction or update →
Benchmark evidence preview

Open original ↗