I have a 10-minute podcast video where the subject’s face is out of focus due to a focus issue during recording.
Previously, I was able to improve similar footage using AI enhancement tools, but the results on this particular video have not been successful. I have already tried CapCut AI Enhancement and AnyEnhancer, but neither was able to recover enough facial detail.
Before purchasing Topaz Video AI, I would like to know if this type of footage is realistically recoverable using your software, especially with features such as Recover Faces and Proteus.
The footage is well-lit, the camera is static, and there is no significant motion blur. The main issue is that the subject’s face is out of focus.
I can provide sample frames or a short clip for evaluation if needed.
There are a few different models and options for workflows here. In some cases using the Focus Fix tool with the enhancement models Proteus, Rhea or Iris can help. However, depending on the severity of the focus issue, they can struggle to lock onto just the facial details.
Another option is the new Starlight Precise 2.5 model as that has shown to be very useful when a video is out of focus.
Here is a quick video that covers some of the options here, if you want to check it out before signing up for the app. We do offer a 48–hour refund window if you want to give it a test as well.
Uh not Rhea What I can offer is to make a few short encodes with different models for you if you provide a source clip with a few seconds of video. I think Starlight Precise could work here, unless the source is noisy or has a pattern that doesn’t suit, in which case you get a “dirty face” effect. Or if there are simply too few pixels, SLP denoised, the few facial pixels become more visible, which also doesn’t look good then. It’s really a matter of experimenting.
if file size is small, you can zip the video and upload it here. You can lossless cut a few seconds from the video, for example with Avidemux. Or use a upload/file transfer portal. I could use Google Drive, but with that I can only offer downloads. Anyone who wants to upload files has to sign in with a Google account, and most people don’t want to do that.
The month-to-month subscription on the product pricing page for either Topaz Video by itself or the full Topaz Studio bundle will be what you want. This will give you access to the desktop app for processing, and you can run the files locally. The month-to-month option can be canceled at anytime if you wish, the other options require an annual commitment, but do come with a discount.
Make sure that your system meets or exceeds the System Requirements for Topaz Video and if you wish to run Starlight Precise 2.5, look at its separate requirements at the bottom of that linked page.
What I can offer is to make a few short encodes with different models for you if you provide a source clip with a few seconds of video. I think Starlight Precise could work here, unless the source is noisy or has a pattern that doesn’t suit, in which case you get a “dirty face” effect. Or if there are simply too few pixels, SLP denoised, the few facial pixels become more visible, which also doesn’t look good then. It’s really a matter of experimenting.
