AI-based Synthetic ID Generation: Creating a Creditworthy Identity from Scratch

[Good Carder]

From carders to carders. Creating a fake identity used to require Photoshop, talent, and a lot of time. In 2026, the barrier to entry has dropped to "write a prompt and get a passport." Generative AI has removed the barriers that have protected identification systems for centuries. For $15 and 5 minutes, you can get a realistic ID card capable of bypassing checks on crypto exchanges, banks, and payment systems. Deepfake videos solve the problem of liveness detection, and LLMs fill in missing pieces of a biography. A synthetic identity is more than just a fake. It's a fully-fledged digital person, capable of taking out loans, opening accounts, and disappearing without a trace.

In this article, I'll analyze how synthetic ID generation works in 2026: from document creation to bypassing biometrics, from building a credit history to automating the entire process. No security — just how it's done in practice.

Part 1: Why 2026 Was a Turning Point for Synthetic ID Fraud​

1.1. Collapse of Barriers​

Three factors converged at one point:

1. Availability of generative models. An audit of 16 leading AI models found that 92% generate fake government IDs on request. Google Gemini, ChatGPT, Grok, and Imagen 4 Ultra create passports, driver's licenses, and national IDs that are visually indistinguishable from the real thing.
2. Speed and price. A convincing fake ID is generated in minutes , for about $15. The OnlyFake platform (a Ukrainian carder) has generated over 10,000 fake documents for clients worldwide.
3. Scalability. AI never tires. A single carder can create thousands of synthetic identities in a day. Traditional KYC systems simply can't keep up with this influx. As of 2026, deepfake attacks account for 11% of global fraud activity, and synthetic ID fraud has grown by 378% year-on-year.

1.2. Synthetic Personality 2026: What is it?​

A synthetic identity isn't just a fake passport. It's a fully-fledged digital human being, complete with:

• An AI-generated ID card with the correct fonts, holograms, and microtext.
• Income verification. AI-generated bank statements, employment certificates, and HR letters.
• Active profile. Deepfake video for live verification with synchronized lip movements and voice.
• Digital footprint. Social media accounts, purchase history, email correspondence — all of this is created automatically.

Part 2. Document Generation: From Passports to Driver's Licenses​

2.1. Which models will work in 2026?​

An audit of 16 commercial AI models found that the following models produce IDs that are convincing to a trained human reviewer:

Model	Ability to generate ID	Quality
Google Gemini (Nano Banana)	High	High (high-fidelity)
ChatGPT (Images 2.0)	High	High
Grok	High	High
Image 4 Ultra	High	High
Recraft v4	High	High

Some models created fake child IDs through standard user interfaces. This suggests that censorship in the models is either absent or easily bypassed.

2.2. Document generation technique​

Step 1. Create a template. Find a publicly available sample of a real document (passport, driver's license) from the desired country.

Step 2. Prepare a prompt. The prompt should be as detailed as possible:

Generate a high-fidelity front-facing digital scan of a [US/UK/EU] driver's license.
Include:
- Full name: [name]
- Date of birth: [date]
- License number: [state-appropriate format]
- Expiration date: [date]
- Photograph: [description of appearance]
- Security features: hologram, microprint, UV patterns, correct typography
- Format: 300 DPI, JPG/PNG
- No watermarks or artifacts

In April 2026, such prompts were circulating openly on the X platform.

Step 3. Post-processing. The generated image may contain micro-artifacts. These are removed in Photoshop or GIMP: fonts are adjusted, scanning noise is added, and the natural aging of the document is simulated.

2.3. OnlyFake: a turnkey solution​

OnlyFake became the most well-known platform for generating fake IDs. The platform allowed:

• Generate driver's licenses for all 50 US states
• Create US passports and passport cards
• Generate IDs from approximately 56 other countries
• Upload your photo and information for personalization

The price is about $15 per document. For this price, you receive a document that passes several online verification systems.

Part 3. Deepfake Video: Bypassing Live Verification​

A static passport photo is no longer sufficient. Almost all KYC systems require selfies with live action. This is where deepfake videos come in.

3.1. Architecture of deepfake attacks​

A modern attack on liveness detection looks like this:

1. Face generation. A synthetic face is created (using StyleGAN, Midjourney, and Stable Diffusion).
2. Video creation. The face is superimposed on a reference video of a real person using DeepFaceLab or ROPE.
3. Voice synchronization. The voice is cloned via ElevenLabs or RVC and synchronized with lip movements.
4. Live spoofing. Using OBS Virtual Cam or similar software, video is broadcast as a "webcam" during live verification.

Key point: liveness detection systems analyze micromovements, blinks, and eye reflections. Modern deepfake models take this into account, creating videos with natural micromovements.

3.2. Tools for deepfake attacks​

Tool	Purpose	Complexity
DeepFaceLive	Real-time face swap via webcam	Average
ROPE	One-shot face swap on video (inswapper_128.onnx)	Low
ElevenLabs	Voice cloning for synchronization	Low
OBS Virtual Cam	Substituting the video source at the system level	Low
DeepFaceLab	Training a model for high-quality replacement	High

3.3. Bypassing Liveness Detection: Step-by-Step​

Step 1. Collecting a reference video. You record (or purchase on the darknet) a video of a person performing typical KYC actions: head turns, blinks, and reciting numbers.
Step 2. Generating a synthetic face. Create a face that will "belong" to your synthetic self.
Step 3. Face replacement. Using ROPE or DeepFaceLab, replace the face in the reference video.
Step 4. Substituting the video stream. Launch OBS Virtual Cam as a video source for the browser. The KYC system sees the deepfake video instead of the real camera.

Important: liveness detection systems look for artifacts in the image. The higher the video resolution and FPS, the lower the chance of detection. Use 1080p video, 30 FPS, with natural noise.

Part 4. Building a Creditworthy Identity: From Zero to a Loan​

A generated document is just the beginning. For an identity to truly "work," it needs a credit history.

4.1. Assembling a Synthetic Personality​

A synthetic personality is a combination of real and fictional data:

• Real SSN/ITIN. Taken from leaks (often the SSNs of deceased people who are not yet listed as dead in databases).
• A fictitious name, address, and date of birth. Generated through Faker or taken from another leak.
• Generated document. Customizable to your name and date of birth.

As experts note, modern fraudsters are using GenAI across the entire synthetic ID pipeline: from document creation and deepfakes to building biographies and social media accounts.

4.2. Building a credit history​

A single document doesn't make you creditworthy. To get a loan, you need a history.

The "build then burn" approach:

1. Register with small services. Open accounts with several neobanks (Revolut, Wise, Payoneer) using a synthetic identity. Make small transactions (purchases, transfers) over the course of 2-3 months.
2. Getting a credit card. Apply for a credit card with a low limit ($500–$1,000). Use it for small purchases and pay on time.
3. Limit increase. After 3-6 months, you request a limit increase. By this point, you already have a payment history.
4. A large loan. You take out a loan of $5,000–$10,000. You cash it out. You disappear.

Important: a synthetic identity created from scratch has no credit history. Therefore, the first steps are to "warm up" the account, not to make an instant cashout.

4.3. Automation via Agentic AI​

By 2026, the process of creating synthetic personalities will be fully automated. AI agents:

• Generate unique identities (first names, last names, usernames, dynamic emails)
• Create documents for each individual
• Registration stages are underway on the platforms
• Manage multiple accounts simultaneously

A single carder with an AI agent can create hundreds of synthetic personalities a day, each of which looks like a real person.

Part 5. Errors in Generating Synthetic IDs and How to Fix Them​

5.1. Error: AI generation artifacts​

Symptom: The document has noticeable unnatural fonts, blurred edges, odd shadows, and incorrect element placement.
Correction: Use high-resolution models (Imagen 4 Ultra, Grok). After generation, process the image in Photoshop: add scanning noise, apply a slight blur, and adjust the colors to match the actual document. Modern AI detection systems specifically look for generation artifacts. The more you "age" the document, the lower the chance of detection.

5.2. Error: Data inconsistency​

Symptom: The name on the passport doesn't match the name on the bank account, and the date of birth doesn't match the age in the photo.
Solution: Maintain a single database for each synthetic identity. All documents, emails, phone numbers, and accounts should be linked to a single profile. Use data generators (like Faker) with the same seed for each identity.

5.3. Mistake: No Digital Footprint​

Symptom: The KYC system doesn't detect social media accounts, purchase history, or email activity.
Solution: Create a minimal digital footprint for each synthetic identity: register accounts on Facebook, Instagram, and LinkedIn; send 1–2 emails per month; make small purchases on Amazon. Experts note that fraudsters are using GenAI to construct biographies with social media accounts.

5.4. Error: Cashout too fast​

Symptom: The persona was registered yesterday, and today they're already trying to get a $10,000 loan.
Solution: "Warm up" the persona for at least 3-6 months. Start with small transactions and gradually increase the limits. The credit history should look like that of a real person, not a bot.

5.5. Error: Data Reuse​

Symptom: The same SSN or address is used for multiple synthetic identities.
Solution: Each identity must be unique. Use different combinations of real and fictitious data. Don't duplicate SSNs, addresses, or phone numbers. Systems link accounts using shared data — one mistake can destroy the entire network.

5.6. Error: Using recognizable AI faces​

Symptom: The synthetic face is too "perfect," symmetrical, and lacks natural asymmetries.
Solution: Use StyleGAN3 with post-processing to add natural imperfections: moles, eye asymmetry, wrinkles. AI-generated faces often have micro-artifacts (incorrect number of teeth, odd ears). Correct these manually.

Part 6. Step-by-step guide: creating a synthetic personality from scratch​

6.1. Infrastructure preparation​

• Anti-detect browser (Dolphin Anty, Octo) for each account.
• Resident proxy (country = country of the document).
• Temporary email and phone (SMS-activate).
• Access to AI models (ChatGPT, Gemini, Grok).

6.2. Personality Generation​

1. Generate data. Use Faker or ready-made scripts to create a name, date of birth, and address.
2. Create a document. Generate a passport or driver's license through ChatGPT or Gemini.
3. Create a selfie. Use Midjourney or Stable Diffusion to generate a photo suitable for documents.
4. Create a deepfake video. Use ROPE or DeepFaceLive to prepare the video for live verification.

6.3. Personality warm-up​

1. Register with neobanks. Revolut, Wise, Payoneer – minimal KYC.
2. Small transactions. Purchases of $10–50, transfers between your own accounts.
3. Credit card. Apply for a low-limit card ($500).
4. Payment history. Pay on time for 3-6 months.

6.4. Cashout​

1. Limit increase. Request a boost to $5,000–$10,000.
2. Large loan. Take out a loan or make a large purchase.
3. Cashing out. Via a crypto card or P2P exchange.
4. Disappearance. Close accounts, destroy documents, change proxies.

Part 7. OPSEC for Synthetic ID Carder​

7.1. Insulation​

• Each synthetic personality is a separate anti-detection profile, a separate proxy, a separate email.
• Never mix data between individuals.
• Don't use a single AI model to generate all documents – rotate models.

7.2. Camouflage​

• Use generation with different seeds for each document.
• Add natural imperfections: scanning noise, slight defects.
• Don't create "perfect" documents - they look suspicious.

7.3. Speed control​

• Don't register 10 identities a day. Spread out the process.
• Don't try to get a loan on the first day after creating your identity.
• Let the individual "live" for at least 3-6 months before a large cashout.

Summary​

Synthetic ID generation in 2026 isn't a hacking skill, it's an industry. 92% of AI models create fake government IDs on request. Deepfake videos bypass live verification. Synthetic identities take out loans, open accounts, and disappear without a trace. The entry threshold is $15 and 5 minutes.

The main errors: AI generation artifacts, data inconsistencies, lack of a digital footprint, and too-quick cashouts. Fix them, and the synthetic identity will last for years.

A quick one-line reminder:
"92% of AI models create passports on request. $15 and 5 minutes — and you're a new person. Deepfake bypasses cameras, SSNs from leaks create credit histories. Warm up the identity for 3-6 months. Don't duplicate data. And never use a perfect AI face — add noise and wrinkles." A synthetic identity is not a fake, it's a digital human being.

Trusted services for creating any document can be found in this section:

Debit Cards, Documents, Tickets, Labels

Any ID, Scans, Holograms, etc

carder.pw