While artificial intelligence technologies have useful aspects, they do for evil purposes It can also be a vehicle. Especially fake videos and photos taken with someone’s face or fake voice recordings made with cloned voices make it very difficult for us to distinguish between real and fake on the internet.

Well, what you see all over the internet How do you know if a video, photo or audio is fake? Unfortunately, the answer to this question is not so simple. However, we will share with you tips that will help you distinguish between real and fake.

The ground rules before we get into everything:

• Be skeptical, ask, don’t sell yourself to the sights and sounds.
• Don’t immediately believe every image you see on the internet.
• Don’t immediately fall for the description images of a loved one that are positive for you.

How to detect fake videos?

• An example from politics, but let’s emphasize this: it won’t always be as easy to detect as in this video.

Pay attention to the person’s facial and body movements:

Of course it’s a mime, the mouth movements are too artificial. But an example nonetheless.

One of the weakest points of deepfake technology, our most human selves. In the images you see, carefully follow the person’s eye, mouth, facial expressions, head and body movements. In this process, look for answers to the following questions:

• Is the person’s skin too smooth or too wrinkled?
• Do the eyes and eyebrows move naturally?
• Does the student stay still or is he playing erratically and randomly?
• Does the person blink a lot or not at all?
• Do mouth movements match the sound you hear? Does the mouth move naturally?
• If so, do the facial beards stay fixed, move a lot or fade?
• Are the body movements natural, do head and body move in harmony?
• Do the impersonators play randomly regardless of what is said?
• Do the emotions and facial expressions conveyed match each other?

Pay attention to the person’s posture and physique:

Another element that is most unsuccessful in artificial intelligence and manual editing is in fake videos human posture and physique. Gently open your eyes and answer these questions:

• Are the person’s body lines natural? Is there a disorder?
• Do the head and body measurements match?
• Does the person adopt a natural posture, especially parallel to the emotion he/she is conveying?
• If there are images from different angles, do the body proportions in these images match the other images?
• Does the skin color match the skin color of the person you see in real images you know and have seen before?

Note the anomalies in the video:

• A study conducted when deepfake went viral.

Humans and artificial intelligence cannot do everything perfectly. may appear in the video. factors such as image distortion, slipping, jammingmay indicate that the video is fake.

• Is there any distortion on the body or face in any frame of the image?
• Does even a pixel that isn’t there suddenly appear and disappear?
• If so, do the items next to the person always remain stable?
• If so, do the items that should be movable move?
• If so, do moving objects move naturally?
• Are there momentary stutters in the video, points that don’t match the previous frame?

Note the accessories on the person, if any:

Artificial intelligence technologies can quite successfully create accessories in fixed images. But with videos it gets harder. Doesn’t know how an accomplice should act or behave at that timeso it usually leaves this accessory to the person.

So try to answer these questions:

• If so, do the glasses naturally fit the person’s eyes?
• Is the light from the glasses reflected too little or too high?
• Does the reflection change or remain fixed as the person’s head moves?
• Do accessories like necklaces, bracelets, and earrings seem to stick or move naturally?
• Does the light reflected from these accessories match the material of the accessory?

Note the lights and shadows in the video:

One of the most successful deepfake videos in terms of adaptation, but mouth movements give it away.

• Do the shadows, especially on the person’s face, appear natural?
• Is the skin color change natural due to lighting, is there much or little difference in different regions?
• Does the lighting change at any point in the video?
• Is there any distortion, even temporary, in the shadows?

How to detect fake photos?

One of the hardest things to detect in the internet world is spotting fake photos. With the ever-evolving talent of technology and digital artists, creating realistic and fake images is just a click away.

Fake Papa images produced with artificial intelligence, such as Midjourney, which produced a scratch-free image even went viral around the world assuming it was real. Unfortunately to distinguish these images You’ll have to be more skeptical.

• Note the person/object borders. Does it fit your entire photo, are there any pixel defects?
• Do the resolutions of the objects, the environment and the person match? Is there anything clearer/blurred?
• Do the shadows match the ambient light? Are there inconsistencies?
• Are the contrast in the environment and the contrast in the person compatible with each other?
• Do the colors in the photo match the person?
• Is there a reflection missing in the environment?
• Is the person’s reflection where it should be? (Mirror, glass, glass, metal surface, polished table etc.)
• Is there anything in perspective that seems strange to you?
• Is there even one deviation in the photo? (There may even be an extra hand on the shoulders)
• Are the person’s feet, torso, and head natural?
• Is there a deterioration in the environment behind the person?

Also for detecting objects added to an image In Adobe Photoshop, you can increase the contrast, play with the color levels, increase/decrease the brightness, or play with the level settings. You may see something like the following:

• This photo won a prize in the Nikon Photography Contest and was later revealed to have been created using Photoshop after several users manipulated the photo.

The following questions can be used to identify images that may have been created with artificial intelligence:

Also, the bend in the index finger is very steep and sharp.

• Are the hands natural?
• Do the limbs and body match the actual images of the person?
• Do the measurements of body and limbs match the actual images of the person?
• Most importantly, does the image match the person’s personality, philosophy, thoughts, style in previous images?

How to detect cloned sounds?

• An example of sound cloning.

Detecting cloned sounds is possible these days, as long as you are careful. Artificial intelligence may not be able to convey the emotion that a person wants to convey in the same way as the cloned person. Even the most advanced cone tools Uberduck and ElevenLabs have some drawbacks.

Look for answers to these questions to detect artificial sounds:

• Do the words spoken by the person and the emotion to be conveyed with these words match the tone of the voice?
• Do you hear a robotic distortion in the sound, even if only for a moment?
• Is there a tone, spike, or dip in the audio that doesn’t match the person’s descriptions you’ve heard before?
• Do the words the person utters match what they have said and said in the past?
• Is the syntax and connection of these sentences close to the way you know the person, is it natural?
• Does the person occasionally breathe while speaking? Is the breathing rate natural according to the speech and what is being said?

Voice cloning technology cannot be done in Turkish with many tools. However, tools are available. However, ask additional questions such as:

• Does speaking Turkish sound natural?
• Does the person’s accent match their previous speech?
• Are the places where the person swallows and breathes natural?
• Does the word stress correspond to the emotion to be conveyed?

Here are some free tools to help you identify fake videos:

If you have any doubts about any of the questions above, you should try to treat the video as fake and not directly believe its content. The truth will soon come to light.

• Deepware
• DuckDuckGoose tool

No free tools have yet been released to detect cloned voices.

​