Hey all,
I've been following deepfake tech since it began and mostly lurk on the forums but post every now and then. I'm not super into the more complicated software and mechanics behind DeepFaceLabs or anything that requires training, so I've more closely followed the image swaps, face swaps for gifs, and ultimately face swaps for videos. Today, we're going to be comparing the available programs/software/web services for the PC that specifically utilize Single-Image Deepfake (SID) technology. The criteria for apps to be considered is that they must be something you can do on your PC (not looking at phone apps), and there cannot be a filter in place that prevents you from uploading explicit content. The following apps that have been reviewed are:
For this test, I used the same set of clips to keep everything equal. In this case, Lana Rain was chosen as the model (cosplaying as Aerith), and I used FF7 Remake's voice actress as the face / model to deepfake.
[Face]
[Comparisons - Open Image For Full-Size]
[img=1289x720]
SimSwap
Pros:
DeepSwap
Pros:
WannaFake
Pros:
FaceHub
Pros:
TL;DR
After analyzing all of these programs, here is my breakdown:
DeepSwap.ai will absolutely give you the best overall results. True 720p output with a high resolution face to blend in nicely. The face is sharp and clear when facing the camera, but can sometimes get a little distorted if at an odd angle or the video is panning too quickly.
SimSwap is a great option for testing what videos will work well with faces before switching to a paid application for better results. As a final product on its own though, it will produce lower quality and often blurry fakes. It's free though, so there's certainly no risk and I believe it's worth trying as an economy option.
FaceHub.live isn't worth the money, at least not against the current competition. Even though it produces some fairly good results, I think the output has too many post-processing type filters applied. The end result does not look nearly as high quality in its 720p video. While the odd angles or action shots look a little better than DeepSwap, I think the straight-on shots (which is what you should be going for anyway) aren't nearly as crisp. Additionally, queue times seem to be ridiculous to process an image.
Wannafake.com is complete garbage. It produces results that are far inferior to the rest of the applications, including SimSwap, which is free. No reason to spend any money on this in my opinion.
**NOTE: You'll notice none of the links that I shared are referral links. This review took quite a bit of time and I ran far more tests on my end than what's shown here to make sure that my results were consistent. If you have any questions about my experience, feel free to ask me. Happy faking!**
I've been following deepfake tech since it began and mostly lurk on the forums but post every now and then. I'm not super into the more complicated software and mechanics behind DeepFaceLabs or anything that requires training, so I've more closely followed the image swaps, face swaps for gifs, and ultimately face swaps for videos. Today, we're going to be comparing the available programs/software/web services for the PC that specifically utilize Single-Image Deepfake (SID) technology. The criteria for apps to be considered is that they must be something you can do on your PC (not looking at phone apps), and there cannot be a filter in place that prevents you from uploading explicit content. The following apps that have been reviewed are:
For this test, I used the same set of clips to keep everything equal. In this case, Lana Rain was chosen as the model (cosplaying as Aerith), and I used FF7 Remake's voice actress as the face / model to deepfake.
[Face]
[Comparisons - Open Image For Full-Size]
[img=1289x720]
SimSwap
Pros:
- Free
- Can be used offline with a good PC
- Cloud version available in case you have a crappy PC or can't figure out how to set it up
- Can output in HD
- Can upload any length of video (though I found that it typically freezes if you exceed 10 minutes)
- Does a decent job at fakes, the blurryness helps "action shots" or odd angles look a little less jarring
- No interface and can be a little challenging to learn and set up for beginners
- Takes a fairly significant amount of time to process (On average 15 minutes for a 2 minute video)
- Requires a lot of processing and graphics power, slows your computer down while it's running
- Face is significantly lower res than the video and tends to be a little blurry
DeepSwap
Pros:
- Incredibly easy to use with a nice interface
- Fantastic faceswap results, usually resulting in a clear face (if using a high quality face image, it should match a high resolution video)
- Outputs to HD (only 720p) but that also helps prevent faces mismatching in resolution quality
- Maximum video length is 2 minutes
- High resolution 2 minute videos process in around 90 seconds.
- Good subscription model (1 month for $10 or 12 months for $50). Provides you 20 credits per month (15s per credit). Additional credits can be purchased if desired.
- Accepts PayPal
- No demo
- Costs money
- Face can sometimes get a little jumbled up when at odd angles (see the angled pic above)
WannaFake
Pros:
- Free 15 second trial video (they used to offer a free 2 minute trial)
- Easy to use with a simple interface
- Costs money (buy time: $70 for 1 hour, $13 for 10 minutes, or $3 for 2 minutes)
- Does not accept PayPal
- Video quality is compressed and resolution looks sub 720p
- Faces swap is blurry and muddy, looks like some kind of post processing filter was applied to the face
- Maximum video length is 1 minute
- Slow video processing time (around 5 minutes for a 1 minute video)
FaceHub
Pros:
- Free 30 second trial video
- Easy to use with a stylish interface
- 2 minute video length
- Faceswap results are fairly good, action shots and odd angles are somewhat blurred but fairly clean
- Accepts PayPal
- Costs money. Two methods, "Package" and "Subscription" (Package: $5 for 5 credits, $37 for 50 credits, $90 for 150 credits | Subscription: $9 for 10 credits per week, $30 for 50 credits per week, $70 for 150 credits per week). 30 seconds per credit.
- Video quality is compressed and resolution looks sub 720p
- Faces appear to have a post processing filter to help smooth them. This works in action shots or at odd angles, but straight on has less detail than DeepSwap.
- Long video queue times, took me over an hour to get through the queue even with "priority access".
TL;DR
After analyzing all of these programs, here is my breakdown:
DeepSwap.ai will absolutely give you the best overall results. True 720p output with a high resolution face to blend in nicely. The face is sharp and clear when facing the camera, but can sometimes get a little distorted if at an odd angle or the video is panning too quickly.
SimSwap is a great option for testing what videos will work well with faces before switching to a paid application for better results. As a final product on its own though, it will produce lower quality and often blurry fakes. It's free though, so there's certainly no risk and I believe it's worth trying as an economy option.
FaceHub.live isn't worth the money, at least not against the current competition. Even though it produces some fairly good results, I think the output has too many post-processing type filters applied. The end result does not look nearly as high quality in its 720p video. While the odd angles or action shots look a little better than DeepSwap, I think the straight-on shots (which is what you should be going for anyway) aren't nearly as crisp. Additionally, queue times seem to be ridiculous to process an image.
Wannafake.com is complete garbage. It produces results that are far inferior to the rest of the applications, including SimSwap, which is free. No reason to spend any money on this in my opinion.
**NOTE: You'll notice none of the links that I shared are referral links. This review took quite a bit of time and I ran far more tests on my end than what's shown here to make sure that my results were consistent. If you have any questions about my experience, feel free to ask me. Happy faking!**
