Unlimited:Waifu2x (WebGPU In-Browser SR) · Image upscaling
Free open-source, login-free on-device processing gives good 4× detail; a dismissible rewarded-ad prompt may appear on first use.
US-West (California, AS2914)
2048 × 2048 · PNG · Open native file ↗
Subsequent pricing and service checks are shown separately without altering historical scores.
One run does not establish all complex scenes, untested paid maxima or long-term uptime.
Plans & historical pricing
unlimited:waifu2x in-browser AI image upscaler · 2026-10-21
1 plan / billing-period records, all listed; — means that period is not recorded.
In-browser WebGPU/WebAssembly upscaler powered by nunif/waifu2x. Completely free with no account system or paid tiers; offers swin_unet and cunet models, 1x/2x/4x upscaling, and denoise levels running on local device hardware.
Cancellation, sources & verification record
Completely free open-source-powered tool with no accounts or subscription charges.
unlimited:waifu2x official page ↗ · 2026-10-21 Capture 1 ↗
Pricing recorded on the test date · 2026-10-17
100% Free & Open-Source (runs on user WebGPU hardware)
Completely free open-source project (no subscriptions or credits)
Historical context only; not current purchase guidance.
Delivery & access
- Free delivery record
- 2048 × 2048 (true 4x lossless PNG export, supporting arbitrary large-image tiled inference)
- Attached file dimensions
- 2048 × 2048 · PNG
Read directly from the attached file; this does not establish free download availability at that size.
- Watermark
- Watermark-free
- Free allowance
- Unlimited local WebGPU runs with zero daily caps or credit deductions
- Account requirements
- 100% usable without login (runs entirely in the local browser via WebGPU)
- Maximum specification record
- 2048 × 2048 (true 4x lossless PNG export, supporting arbitrary large-image tiled inference)
May include advertised capabilities; this does not establish a tested paid tier.
- Batch & API
- Supports multi-model switching (Swin-UNet Art / Art Scan / Photo), 4-level denoise, and configurable tile size
- Manual refinement
- Not recorded or not verified
Tested result & sample
Waifu2x Unlimited (Swin-UNet Photo 4x) suppresses JPEG compression artifacts cleanly while preserving natural camera gradients (PSNR 28.34 dB, SSIM 0.895) in 2.63s via WebGPU, without introducing fake generative hallucinations.
- Fixture
- anchor_upscale_v1
- Recorded environment
- US-West (California, AS2914)
- Local processing record
- 2.63 s
- End-to-end reference
- 2.63 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
- Explicitly discloses WebGPU browser requirements and model options at the top of the page
- Interaction
- Pure single-page engineering workbench: select model (Swin-UNet Art / Photo), Denoise level, and 1x/2x/4x scale, then immediately preview or download the 2048×2048 PNG.
- Entity record
- nagadomi (waifu2x / nunif Open-Source Author)
- Origin
- Independent open-source AI vision researcher (Japan)
- Domain history record
- 2015–Present (waifu2x.net)
- Third-party tracking record
- Not recorded or not verified
- External information
- Not recorded or not verified
- Processing
- 100% Local Browser WebGPU / WASM ONNX Tiled Inference
- Retention statement
- Zero server upload — all pixel data remains strictly inside local browser memory
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.
anchor_upscale_v1 · Inspect fixture & file inspector →
Captures entry rules, result view and export disclosure.
Screenshots, native output and official claims support different conclusions. Evidence grades are not product safety certifications.
Evaluation basis & full scores · Overall 89.1/100
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_upscale_v1
Honesty & reputation · 92.5/100 · 30%
| Criterion | Score | Rationale | Basis |
|---|---|---|---|
| Pricing transparency | 34 / 35 | 100% free and open-source in-browser WebGPU tool with zero paywalls, zero ad clutter, and no usage limits (01_landing.png, 02_workspace.png, 03_download.png). | Evidence A → |
| Interaction restraint | 25 / 30 | Pure single-page engineering workbench: select model (Swin-UNet Art / Photo), Denoise level, and 1x/2x/4x scale, then immediately preview or download the 2048×2048 PNG. | Evidence A → |
| Background & reputation | 19 / 20 | Authored and maintained by nagadomi, the creator of the iconic Waifu2x open-source repository and the nunif deep-learning framework. | Evidence A → |
| Engineering & privacy | 14.5 / 15 | 100% local in-browser WebGPU / ONNX Runtime inference — user images never leave the local device, offering maximum privacy protection. | Evidence A → |
Capability · 87.7/100 · 70%
| Criterion | Score | Rationale | Basis |
|---|---|---|---|
| Task quality | 57.2 / 65 | PSNR vs 2048px GT: 28.34 dB · SSIM: 0.895 · Edge Gradient Correlation: 0.7165 · High-Freq Sharpness Ratio: 0.5256 · Calibration Plate PSNR: 17 dB | Evidence A → |
| Deliverable specifications | 20 / 20 | Exports a full 2048×2048 4x upscaled PNG (or WebP/JPEG) locally in the browser with zero watermarks and no file-count limits (20.0/20) | Evidence A → |
| Processing speed | 10.5 / 15 | Local WebGPU hardware acceleration completes 512→2048 (4x) tiled inference in just 2.63s with zero network upload overhead (14.5/15) | 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
- 2026-10-17
Initial real-machine local WebGPU 4x benchmark on Mode A ground-truth anchor (anchor_upscale_v1).