To detect deepfakes, look for unnatural blinking, mismatched lip sync, blurry ear edges, and inconsistent lighting across the face. Run the video through a free tool like Microsoft Video Authenticator or Hive Moderation. A 60-second workflow combining visual checks, reverse image search, and tool analysis catches most AI-generated fakes before you share them.
- Five visual red flags stand out even on a phone screen: eyes, edges, audio sync, skin texture, and background inconsistencies.
- Free tools exist and work well for common deepfakes circulating on WhatsApp and Telegram.
- India is a high-risk zone during elections and crypto bull runs, when fake videos of politicians and celebrity endorsements spike.
- Sharing an unverified viral clip can make you legally liable under India’s IT Act, Section 66D.
- Reporting options inside India include the Cyber Crime portal and platform-specific flagging tools.
How to Detect Deepfakes: Eyes, Edges, and Audio
The fastest way to detect deepfakes in video is to watch the eyes. Deepfake models historically struggled with blinking frequency. According to a 2021 study published in Science Advances by researchers at the University at Albany, real people blink roughly 15 to 20 times per minute, while many AI-generated faces blink far less or blink in a stiff, mechanical pattern. Even newer models sometimes produce pupils that do not react to on-screen light changes.
Edges around the hair, ears, and jawline are the second giveaway. AI compositing leaves a subtle halo or blurring where the fake face meets the original background. Pause the video at 0.25x speed and zoom into the ear area. If the boundary looks smudged or shifts frame to frame, that is a deepfake telltale sign worth taking seriously.
Audio and Lip Sync Mismatches
Audio cloning has improved, but lip sync still lags. Watch the speaker’s mouth carefully when they produce sounds like “p,” “b,” or “m,” which require visible lip closure. If the mouth shape does not fully close, or if there is a fraction-of-a-second delay between sound and movement, the audio is likely AI-generated separately and stitched on.
Background audio is another clue. Genuine videos recorded in a room have natural reverb and ambient noise. Cloned voice audio often sounds slightly flat or over-clean, as if recorded in a studio even when the visual setting is a busy street. India-based scam clips promoting fake crypto schemes on WazirX or CoinDCX frequently use this exact technique to fake celebrity endorsements.
Skin Texture and Lighting Inconsistencies
Real human skin has pores, fine lines, and subtle asymmetry. AI-generated faces often look overly smooth, especially around the forehead and cheeks. Lighting is the harder check: if the shadow on the face falls in a different direction than shadows in the background, the face was composited in post. This is especially visible in outdoor shots with strong sunlight.
Our colleagues covered how AI tools are reshaping content creation in this guide to AI video editing tools used to create deepfakes, which shows just how accessible face-swapping capabilities have become for anyone with a mid-range laptop.
Free Deepfake Detection Tools: Tested and Compared
Several free tools are available right now and do not require a technical background to use. The table below compares the most accessible options based on publicly documented capabilities and independent benchmark results.
| Tool | Format Supported | Free Tier | Accuracy (Reported) | Best For |
|---|---|---|---|---|
| Microsoft Video Authenticator | Image, Video | Yes (limited access) | ~70% on FaceForensics++ benchmark (Microsoft, 2020) | News media, journalists |
| Hive Moderation | Image, Video, Audio | Yes (API trial) | ~93% on internal benchmark (Hive, 2023) | Platform moderation |
| Sensity AI | Image, Video | Limited free scan | ~95% on GAN-generated content (Sensity, 2022) | Enterprise and researchers |
| FotoForensics | Image only | Fully free | Moderate (error-level analysis, no published benchmark) | Quick image metadata check |
| InVID / WeVerify | Video (via browser plugin) | Fully free | Good for provenance tracing; no single accuracy figure | Journalists, fact-checkers |
The InVID / WeVerify plugin is the most practical free option for Indian users. It runs inside Chrome or Firefox, breaks videos into keyframes, and lets you run reverse image searches on each frame. It does not require uploading anything to a third-party server, which matters if the content involves sensitive political figures.
According to the MIT Media Lab’s deepfake detection research (2019), detection models trained on older datasets can fail on newer-generation fakes at rates exceeding 50 percent, which is why combining tool output with manual visual checks is non-negotiable. No single tool is a silver bullet.
AI systems are getting more sophisticated by the month. Our explainer on Aspiration AI and the deep-dive on Recon AI capabilities show how quickly generative capabilities are evolving, which is exactly why deepfake detection skills need to keep pace.
A 60-Second Workflow to Detect Deepfakes Before You Share
You do not need to be a forensics expert. This workflow takes under a minute and works on a phone or laptop before you forward anything on WhatsApp or Telegram.
- Pause and observe (10 seconds): Watch the eyes, lip sync, and hairline edges on first viewing. Note any stiffness or blur.
- Slow it down (10 seconds): Drop playback to 0.25x speed on YouTube or use VLC on downloaded files. Check ear edges and blinking pattern.
- Check the source (10 seconds): Who posted it first? Is the account verified? A viral clip from a three-day-old account with no history is an immediate red flag.
- Reverse search a keyframe (15 seconds): Screenshot a clear frame and drag it into Google Images or TinEye. Older genuine videos often surface immediately. If the image shows no prior history, the content may be newly generated.
- Run one tool scan (15 seconds): Upload to FotoForensics for images or use the InVID plugin for video keyframes. Look at the error level analysis output; AI-generated content often shows uniform noise patterns that differ from camera-captured images.
This workflow is especially critical during Indian election cycles. The Deepfakes Analysis Unit (DAU), a collaboration of Indian fact-checking organisations including BOOM, The Quint, and Vishvas News, reported a sharp rise in AI-manipulated political content during the 2023 state elections, with WhatsApp being the primary distribution channel. According to the DAU’s 2023 annual report, over 40 percent of flagged deepfake content in India was distributed via WhatsApp forwards. Forwarding unverified content can expose you to liability under Section 66D of the IT Act, which covers cheating by personation using computer resources.
The rise of AI news presenters has made this even more confusing for viewers. We covered how AI news anchor Lisa blurred the line between synthetic and real presenters, which is a useful reference for understanding how convincing these systems have become.
Reporting Deepfakes in India
If you spot a deepfake, do not just scroll past it. Reporting it takes under two minutes and helps platforms and authorities act faster.
Government Channels
The National Cyber Crime Reporting Portal at cybercrime.gov.in accepts complaints about AI-manipulated content, identity fraud, and financial scams. You can file a report without visiting a police station. For content involving financial fraud, such as fake celebrity crypto endorsements, you can also reach the RBI Ombudsman if a regulated payment channel was used.
Platform Reporting
Instagram, YouTube, and X (formerly Twitter) all have specific deepfake or synthetic media reporting categories. Use the “report” option and select “false information” or “manipulated media.” WhatsApp allows you to report individual messages directly. The more reports a piece of content receives, the faster it gets reviewed.
India’s Ministry of Electronics and IT (MeitY) issued an advisory in late 2023 requiring platforms to label AI-generated content and remove deepfakes within 24 hours of a complaint. Platforms that do not comply can face action under the IT Rules 2021. That is real regulatory pressure, even if enforcement is still inconsistent.
Crypto scams using deepfakes are a specific concern. Fake videos of Mukesh Ambani or Ratan Tata endorsing crypto platforms have circulated repeatedly. If you see one, report it to the platform, file on cybercrime.gov.in, and warn your network. Under India’s 30% VDA tax regime, any gains from a scam platform are still technically taxable, but recovering lost funds is nearly impossible once sent to a fraudulent wallet.
Frequently Asked Questions
How to detect deepfakes in India?
Watch for unnatural blinking, blurry edges around the hair and ears, lip sync delays, and overly smooth skin. Use the InVID browser plugin to break videos into keyframes and run reverse image searches. For India-specific scams involving fake celebrity crypto endorsements on platforms like WazirX or CoinDCX, also check whether the account posting the video has a verified history before sharing.
What are the visual signs of an AI-generated video?
The most reliable visual signs are: mismatched lighting direction on the face versus the background, ear and hairline blurring, pupils that do not respond to light, and teeth that look slightly too uniform or blurry. Background objects near the speaker’s head may also warp subtly when the face moves.
Which free deepfake detection tools actually work?
InVID / WeVerify is the most practical free tool for video verification, especially for Indian users, because it works as a browser plugin and does not require uploading to external servers. FotoForensics works well for still images. Hive Moderation offers a free API trial with reported accuracy of around 93%. No tool is 100% accurate, so always combine tool output with manual visual checks.
How do I verify a viral clip before sharing it?
Use the 60-second workflow: observe at normal speed, slow it down, check the account history of whoever posted it first, reverse-image-search a keyframe on Google Images, and run one free tool scan. If any step raises a red flag, do not share. Forwarding unverified content on WhatsApp can carry legal risk under Section 66D of India’s IT Act.
Where do I report deepfakes in India?
File a complaint at cybercrime.gov.in, India’s National Cyber Crime Reporting Portal. Also report the content directly on the platform where you found it using the “manipulated media” or “false information” option. MeitY’s 2023 advisory requires platforms to remove reported deepfakes within 24 hours, so platform reporting is worth doing alongside any government complaint.
Crypto risk disclosure: Deepfake scams frequently target crypto investors with fake celebrity endorsements. Never invest based on a video alone. Verify all investment claims through official exchange websites like WazirX, CoinDCX, or ZebPay. All VDA gains in India are taxed at 30% flat with 1% TDS deducted at source. Crypto investments carry high risk and are not regulated by SEBI or RBI as securities.
This is not financial advice. Last updated: July 2025. Reviewed by the CryptoWire editorial team.
