Terms of Service
Last updated: September 30, 2026
1. Acceptance of terms
Welcome to Feyard. These Terms of Service (“Terms”) govern your access to and use of the browser-based AI tools and related services operated by Weiqing Li, an individual operating under the laws of China (“Feyard”, “we”, “us”, or “our”), available at feyard.com (the “Service”).
By creating an account, subscribing to a paid plan, making a purchase, or otherwise using the Service, you confirm that you: (a) are at least 18 years of age; (b) have read, understood, and agree to be bound by these Terms; (c) agree to our Privacy Policy; and (d) are authorized to enter into this agreement on your own behalf or on behalf of any organization you represent. If you do not agree, please do not use the Service. Continued use after an update constitutes acceptance of the revised Terms.
2. The Service
2.1 What we provide
Feyard provides a collection of browser-based AI tools for image, video, audio and document tasks. The Service is a digital, internet-delivered software product. Access to paid features is granted immediately after payment is confirmed; because of the instant nature of digital delivery, refunds are subject to the specific limits in Section 7.
2.2 On-device processing
Our tools are designed to run locally in your browser using on-device inference technologies (such as WebGPU and WebCodecs). The files you process remain on your own device and are not uploaded to our servers. Some tools download open-source model files to your browser for local execution; such models remain subject to their own licenses.
2.3 Free and paid tiers
All tools are available free of charge with standard processing limits. Paid Pro plans raise selected limits. Current plan details and pricing are published at feyard.com/pricing. Tools marked as planned or coming soon may change or may not ship. We may modify, suspend, or discontinue any part of the Service at any time.
2.4 Underlying AI models
Our tools are not a rebranded third-party API: the open-source models below are downloaded into your browser and run locally via ONNX Runtime / Transformers.js (WebGPU or WebAssembly). Model weights are cached on your device; you can inspect each model at its linked source page. We are independent of these model providers and are not endorsed by them.
| Tool | Model(s) | Origin / publisher |
|---|---|---|
| Image upscaler | Swin2SR (realworld-sr-x4-64-bsrgan-psnr), ONNX Web build; Real-ESRGAN anime6b for illustrations; GFPGAN v1.4 for optional face enhancement | Microsoft Research (Swin2SR); ONNX build by Xenova — model card; Xintao Wang et al. / Tencent ARC Lab (Real-ESRGAN anime6b, BSD-3) — model card; Tencent ARC Lab (GFPGAN, Apache-2.0) — model card |
| Photo restoration | Swin2SR for super-resolution/denoising, GFPGAN v1.4 for face restoration, LaMa for scratch & damage inpainting | Microsoft Research; Tencent ARC Lab (GFPGAN) — GFPGAN model card; Samsung AI Center (LaMa), ONNX build by Carve — LaMa model card |
| Black-and-white photo colorization (Restore) | DDColor-tiny (fp16 ONNX weights, fp32 I/O) | DDColor by piddnad et al. (ICCV 2023, Apache-2.0); ONNX build by edgetools — model card · upstream project |
| Object / watermark removal | LaMa (ONNX build) | Samsung AI Center (LaMa); ONNX build by Carve — model card |
| Magic Select (click-to-segment, Eraser Pro) | SlimSAM-77-uniform (quantized + fp16 ONNX) | SlimSAM is a distilled Segment Anything Model (Apache-2.0); ONNX build by Xenova — model card |
| Background removal | BiRefNet, MODNet | Open-source research models (ZhengPeng7/BiRefNet; ZHKKKe/MODNet ONNX build by Xenova) |
| 3D depth reconstruction | Depth Anything V2 — Small | ByteDance / TikTok (Depth Anything V2); ONNX build by onnx-community — model card |
| Text-to-speech | Kokoro-82M v1.0 (English/multilingual) and v1.1-zh (Chinese) | Kokoro is an open-weight, Apache-licensed speech model in the StyleTTS2 family (Yingfan Aaron Li et al.); ONNX builds published by onnx-community and run locally via kokoro-js — model card |
| Speech-to-text / subtitles | Whisper-tiny, Whisper-base (timestamped) | OpenAI (Whisper); ONNX builds by onnx-community — model card |
| Video background blur | MediaPipe Selfie Segmentation (ONNX build) | Google (MediaPipe); ONNX build by onnx-community — model card |
| Video AI denoise | DTLN, two-stage 16 kHz ONNX | Breizhn — N. L. Westhausen & B. T. Meyer, Interspeech 2020; ONNX weights by niobures — model card |
| Video super-resolution | Real-ESRGAN General x4v3 (SRVGGNetCompact), ONNX build | Tencent ARC Lab (Xintao Wang et al.); ONNX build by tamnvcc — model card |
| PDF summary / text humanizer / diagram generation | SmolLM2-1.7B-Instruct, Qwen2.5-1.5B-Instruct | Hugging Face (SmolLM2); Alibaba Cloud (Qwen2.5), Web/ONNX builds |
| Document Q&A / RAG | all-MiniLM-L6-v2 embeddings | sentence-transformers project |
| OCR | Tesseract (OCR engine); moondream1 (vision-language model) | Google (Tesseract, Apache-2.0); open-source moondream vision model |
| Compression, vectorization, cleaning, puzzles, audio tools | Classical algorithms (codecs, contour tracing, signal processing) — no generative model | Implemented in-house |
| Camera RAW development (RAW Studio) | Classical algorithms (TIFF/ISO-BMFF container parsing, lossless JPEG SOF3 decoding, gradient-corrected demosaicing, Bradford color adaptation, linear-light develop pipeline) — no generative model | Implemented in-house against public format specifications (TIFF 6.0, DNG 1.x, ISO 12234-2 / ITU-T T.81 lossless JPEG) |
Models can be updated as better open-weight alternatives become available; this table is kept current with each release.
3. Accounts
Some features, including paid subscriptions, require an account. You must provide accurate information and keep your login credentials confidential. You are responsible for all activity under your account. Notify us promptly at hello@feyard.com of any unauthorized use. If you sign in with a third-party provider (such as Google or GitHub), you also agree to that provider’s terms.
4. Subscriptions and one-time purchases
We offer the following paid options:
- Pro Monthly — a recurring monthly subscription, currently USD $9 per month.
- Pro Yearly — a recurring yearly subscription, currently USD $79 per year.
- Lifetime — a single one-time payment, currently USD $149, granting Pro access for the lifetime of the Service.
Current pricing is always displayed at the pricing page. We may introduce new plans or change plan features. Price changes for existing subscriptions take effect only on your next renewal period and we will notify you in advance.
5. Billing and payment
5.1 Recurring payment authorization
By providing a payment method and starting a paid subscription, you expressly authorize us and our payment processor to charge the subscription fee to your payment method at the start of each billing period. This authorization remains in effect until you cancel under Section 6. Your subscription renews automatically at the end of each billing period unless you cancel. If a renewal payment fails, we may retry charging your payment method and may restrict Pro access until the payment succeeds.
5.2 One-time purchases
A Lifetime purchase is a single, non-recurring charge authorized at checkout. There are no further payments unless you separately purchase another product.
5.3 Taxes
Prices are exclusive of applicable taxes (including VAT, GST and sales tax) unless shown otherwise at checkout. Where required by law, taxes are collected and remitted on your behalf by our payment processor.
5.4 Payment processing
Payments are processed exclusively by Waffo Pancake, which acts as the merchant of record and a PCI-DSS compliant payment processor. Card details are handled by Waffo Pancake and are never stored on our servers. Charges on your card statement will appear under a Waffo Pancake descriptor; contact billing@feyard.comif you do not recognize a charge.
6. Cancellation
6.1 How to cancel
You can cancel a subscription at any time, in either of the following ways:
- Online (recommended): open the self-service customer portal linked from your Waffo Pancake receipt or billing emails, or email billing@feyard.com from your registered address and we will send you a one-click management link.
- By email: send a cancellation request to billing@feyard.com.
6.2 Effect of cancellation
- Pro access remains active until the end of the current billing period.
- No further charges will be made after cancellation takes effect.
- A canceled account reverts to the free tier automatically when the paid period ends.
- Lifetime purchases are not recurring and require no cancellation; Pro access persists for the lifetime of the Service.
7. Refund policy
7.1 Subscriptions — 7-day guarantee
First-time subscribers may request a full refund within 7 days of their first charge, provided the paid features have not been substantially used. Beyond that window, subscription fees for a billing period you have already used are generally non-refundable.
7.2 Lifetime — 30-day guarantee
One-time Lifetime purchases can be refunded in full within 30 days of purchase if you are not satisfied.
7.3 Always refunded
- Duplicate or erroneous charges caused by a billing error.
- Verified Service outages of more than 72 hours caused by us — a pro-rata refund or extension.
- Refunds required by mandatory consumer law, including the 14-day withdrawal right applicable in the EU/UK where it has not lapsed.
7.4 How to request a refund
Email billing@feyard.com with your account email, the transaction identifier from your receipt, and the reason. We will confirm within 2 business days and, when approved, issue the refund to the original payment method within 5–10 business days. Refunds are not available for accounts terminated for violating these Terms.
8. Billing disputes
If you believe a charge is incorrect, contact billing@feyard.com before initiating a chargeback with your bank. We will respond within 2 business days and resolve confirmed billing errors within 5 business days.
9. Your content and intellectual property
9.1 Your files stay yours
Because processing happens on your device, we never receive, store, or claim ownership of the files you open in our tools or the content you produce with them (“Output”). You retain all rights to your files and may use the Output for any lawful purpose.
9.2 No training use
We do not use your files or Output to train machine-learning models. The open-source models used by the tools run entirely in your browser.
9.3 Output accuracy
AI-generated results may be inaccurate, incomplete, or outdated. You are responsible for reviewing Output before relying on it, and it must not be treated as professional legal, medical, financial, or other regulated advice.
9.4 Our intellectual property
The Service’s design, branding, and code are owned by Feyard and protected by intellectual-property laws. Open-source components remain subject to their respective licenses.
10. Content policy and acceptable use
This section applies to all content created, edited, synthesized, or distributed using the Service — including images, video, audio (including synthetic speech), text, documents, and the inputs you provide to tools.
10.1 Prohibited content — six categories
You may not use the Service to create, edit, enhance, or distribute:
- Pornography and adult (NSFW) content — sexually explicit imagery, audio, or text, whether or not it involves real people.
- Violence and gore — depictions of real-world killing, torture, severe bodily injury, or gore intended to shock, glorify, or incite violence.
- Hate speech and harassment — content attacking, dehumanizing, or inciting violence or hatred against individuals or groups based on race, ethnicity, religion, disability, gender, sexual orientation, nationality, or other protected characteristics.
- Child-unsafe content — any sexualized content involving minors, or any content that harms, exploits, or endangers children.
- Non-consensual deepfakes and impersonation — AI-synthesized or AI-altered images, video, or voice that impersonate a real, identifiable person without consent, including fabricated statements or conduct attributed to them.
- Infringing content — content that violates copyright, trademark, publicity, or privacy rights, or that you do not have the rights to process.
10.2 Zero-tolerance items
The following result in immediate action and may be reported to authorities:
- Child sexual abuse material (CSAM) or any sexual content involving minors.
- Content promoting or facilitating terrorism, extremism, or mass violence.
- Instructions for producing weapons of mass destruction.
- Content inciting genocide or ethnic hatred.
- Non-consensual sexual deepfakes targeting real individuals.
10.3 Other prohibited conduct
- Generating or spreading deliberate misinformation or fabricated claims.
- Passing AI-generated or AI-restored content off as authentic human-created evidence with intent to deceive.
- Bulk-generating spam, phishing content, malware, or deceptive commercial material.
- Circumventing the Service’s security, payment, or access-control features, including unlocking Pro limits without payment.
- Using inputs, outputs, models, or derived data to train, fine-tune, benchmark, or distill competing AI models, or systematically scraping model outputs.
- Reselling or redistributing access without prior written permission.
- Violating any applicable law, regulation, or card-scheme rule.
10.4 How moderation works in this on-device product
Unlike cloud AI services, our tools run entirely in your browser: the files you process and the text you synthesize are not transmitted to, inspected by, or stored on our servers (see Section 2.2). We therefore cannot perform real-time server-side scanning of individual inputs or outputs, and we do not pretend to. The safeguards that genuinely apply are:
- Account-level controls — most tools work without an account; features requiring an account (including Pro) are tied to a verified identity and can be suspended or terminated.
- Manual review of reports — every report under Section 10.5 is reviewed by a human, and where evidence is supplied (such as a screenshot or a link), we assess it against this policy.
- Payment and fraud screening — checkout and payment are screened by our payment processor Waffo Pancake.
- Network-level abuse protection — our hosting provider applies automated security and abuse controls to site traffic.
Because processed files never reach us, members of the public who encounter abusive output elsewhere are our primary detection channel; reports are reviewed promptly and actioned under Sections 10.6–10.7.
10.5 Reporting abusive content
If you encounter content created or edited with the Service that violates this policy — in particular child-safety content, non-consensual deepfakes, hate speech, or content infringing your own rights — report it to report@feyard.com, or via the general support channel hello@feyard.com. Include what you found, where (a link or screenshot is sufficient), and why it violates this policy.
- Reporter identities are kept confidential.
- Every report receives an independent, good-faith review.
- For apparent illegal content involving child safety or terrorism, we will report it to the competent authorities.
- Malicious or knowingly false reports may themselves lead to action.
10.6 Severity and response times
| Level | Examples | Initial response | Resolution |
|---|---|---|---|
| L1 — Critical | CSAM, terrorism, non-consensual sexual deepfakes | Within 24 hours | Within 3 business days |
| L2 — High | Other NSFW, gore, hate speech, targeted impersonation | Within 2 business days | Within 5 business days |
| L3 — Medium | Infringement, deceptive use, general policy violations | Within 3 business days | Within 10 business days |
| L4 — Low | Minor or unclear violations | Within 5 business days | Within 15 business days |
10.7 Enforcement actions
Depending on severity and repetition, we may:
- Issue a written warning and require the activity to stop.
- Restrict access to specific features or reduce usage limits.
- Suspend the account pending review.
- Permanently ban the account and terminate paid access without refund.
- Report illegal activity to the competent authorities and cooperate with investigations.
10.8 Appeals
If you believe an enforcement action was taken in error, appeal within 14 days to legal@feyard.com with the relevant details. Appeals are reviewed independently of the original decision, normally within 10 business days.
11. Data, privacy, and security
Your use of the Service is governed by our Privacy Policy. Payment-card data is processed exclusively by our PCI-DSS compliant payment processor Waffo Pancake and is never stored on our servers. Account data is retained for 90 days after account deletion and then deleted, except where longer retention is required by law.
12. Disclaimers and limitation of liability
12.1 Disclaimer
The Service is provided “as is” and “as available”, without warranties of any kind, express or implied, including merchantability, fitness for a particular purpose, non-infringement, or accuracy of AI-generated output. We do not warrant that the Service will be uninterrupted or error-free.
12.2 Limitation of liability
To the maximum extent permitted by applicable law, Feyard is not liable for any indirect, incidental, consequential, or punitive damages. Our total liability arising out of or relating to the Service is limited to the amount you paid us in the 12 months preceding the claim.
13. Term and termination
These Terms remain in effect while you use the Service. We may suspend or terminate accounts for material breach, fraudulent activity, or where required by law. If we discontinue the Service for reasons other than your breach, we will refund the unused portion of any prepaid subscription period and, for Lifetime purchases, a pro-rata refund reflecting the remaining lifetime of the Service. You may delete your account at any time by contacting hello@feyard.com.
14. Governing law and disputes
These Terms are governed by the laws of China, without regard to conflict-of-law rules. Before initiating formal proceedings, please contact legal@feyard.com to attempt an informal resolution. Nothing in this section limits mandatory consumer-protection rights available to you under the law of your country of residence.
15. General
We may update these Terms from time to time. Material changes will be communicated at least 14 days before they take effect, via notice on the site or email to registered users. Continued use after the effective date constitutes acceptance. If any provision is held invalid, the remaining provisions remain in full effect. These Terms, together with the Privacy Policy, constitute the entire agreement between you and Feyard.
16. Contact
- General support: hello@feyard.com
- Billing, refunds, and cancellation: billing@feyard.com
- Content reports (abusive or illegal AI-generated content): report@feyard.com
- Appeals: legal@feyard.com
- Privacy: privacy@feyard.com
Operator: Weiqing Li, an individual operator in China. Website: feyard.com.