as per Ann, Audition has a tool where you can create a noise print, and attempt to use that to remove the noise (works best if you have a sectrion of audio with no other sounds but the noise you want to remove).  Bottom line, you will have to expirment, it could be really easy or go the other way too.  First step, poke around in there and try some of the tools and see if youtube has a 'how to' for the tools your trying out.   Remove ANY Background Noise from Your Audio File!   How to Remove Background Noise in Adobe Audition 2025   Can I Repair This Noisy Ruined Audio in Premiere Pro!?   How to Isolate and Remove Unwanted Sounds in Adobe Audition CC   to get you started.... ... View more

@JonesVid I am wondering if you have tried the newer noise models mile NYX Large?  I have not tied NEAT Video for a long time now,  but sounds like you have recent experence with both - just wondering if the Large model NYX exceeds NEAT in your opnion.   I have typicaly have good results from BorisFX noise reduction, and rearly use Topaz for just noise reduction.  I find that the upscalling process itself tends to remove all the noise needed and a second noise process wasnt needed - but my expections with this process may have been too low...  despite being happy with it the results. ... View more

I read through that, its a good conversation about a few options for software.  The thread killed off the topaz conversation real quick with what appears to comparison to years old models/versions.   Another person also recommend BorisFX, UP_RES, but that is not a restoration tool and will look like a hot mess with DVD encoded quality as a source (its also just as slow as Topaz with no restoration benefit so not a good tool for this OP's intent).   I think the most efficient way to a good restoration is to get the raw video off the homemade Disc, interlaced and all, and run it through TOPAZ a few different ways, have Topaz handle the De-interlace, and also double the frame rate for the deinterlace.  It will require some testing, which could be done inside topaz with the preview, but if your new to the software, and don't know what t expect with the viable models, your better off have a trimmed 10 second clip, and running it through all the viable models overnight I think.   The topaz forums should have some posts on the process and models people were using, and they have lots of people that do this on the regular that could assist/fast track a process for the OP.   ... View more

@spinachworld some unsolicted advice for you....  as someone who has done this a few times (during covid, I converted, and cleaned up about 200 hours of family videos to burn through the days).   I used Topaz,  and here are my biggest learnings. 1) DO NOT de-interlace until you find a spot in the process where it makes sense,  Handbrake did my deinterlace and I regretted it, as I found better ways in topaz to do it later in the process which much better results. 2) I always kept the original files,  as I was doing it for so long, I found Topaz was updating the tools, or I learned a new process that made it worth going to back to some videos to re-do them. 3) The files pulled off the DVD were of very poor quality compared to HI8, and other first source masters,  the DVD encoding process really did a number onlonger videos back in the day for compression.  Topaz does a really job job removing mpeg compression blocks btw. 4) I would try and do everything in Topaz for clean up, they have some amazing tools now (models)- the Noise removal tools were really good.  Also Starlight is AMAZING in many use cases (not all) - but its so slow if you process it locally, and expensive if you use their cloud service.  I would get myself a small sample clip, of each variant of video quality you have, and run them through a few models, and see what you like.   5) (ADDED LATE)  -  watch the faces!  these are the hardest thing to handle with any AI tools,  check them arefull because they can make the person look like someone else.  I found you had to carefully balance the process because oen AI model can do an amazing job with everything but peoples faces, and if you didnt check, end up with some scary results - so this is a place where you want to have test process fot the models, with faces (as an FYI, faces in the distance with low detail are effected different that faces closer to the camers BTW - this model selection with faces are involved are require the most supervision from you/humans).   Also;  go to the topaz forum,  there are a number of helpfull people on there with this sort of stuff for this software.  There are apperently other alternatives now to Topaz, but I have not needed to try them   As an aside, if you own BorisFX, they have a pretty good upscaler ML_UPRES that works realy good, but the source has to be decent to start with, Topaz is significantly better as the quality degrades.  I Use ML_UPRES when its 720p or 1080p to get to 4k or want to do a zoom.   The process of cleaning up the video will take LOTS of time - be prepared for that - the newest versions of Topaz now give you an estimated time thats pretty close.   Depending on your GPU, you need to decide if your system can handle 1 to 4 jobs at the same time.  I would recomend never goign to 4,  3 might even be too high (but I'm an idiot and I use 3) - I think 2 may be the best,  but test with one, and see how much CPU, System memory, GPU, and GPU memory is left before you increase the number of workloads in the que - if you get this wrong on a small batch, its not too much wasted time - but on a big batch you can add days to the process when it wasnt needed.     ... View more

I finaly got my 5090FE and installed it today;  managed to render out the same project that was about 40 min long in 3 different ways, log file below for details, but here is the specs and versions used for the software for your referance, spoiler, it rendered without any issues.  Id be open to downloading a project someone has issues with, and testing to see if I also run into issues.   Video started as a Iphone HDR H265, converted to PRORES for the first test, and exported back to H265 in 4K   Second test, used the original H265 HDR footage, and exported it from 4K H265 to 1080P (Seq size) H265, also no issues   Third Test, used the original H265 HDR footage, exported it from 4K H265 to 1080p (Seq size) H265, also no issues (I think this matches the settings of the OP, if I read thier post fully/correct.   There was not a single blip, and the 5090 had lots of head room;  I was running the new masking tool in the over all three export tests,  the 5090 doesnt seem to speed up the new mask took in any way, but none the less it was running within premier 25.6 Beta, and also watchign youtube videos at 3x speed.   And now, the important part, Drivers/software versions:   Windows 25H2 Build26200.6584 (This is a REALLY new offical version, if you want it, I can tell you how to trigger the download/install).   Adobe Premier Pro Beta 25.6.0.78 2025-09-23 Media Encoder Beta 25.6.0.074 2025-09-25 (I used Media encoder to export with)   Nvidia Studio Driver 581.29 Sept 10, 2025   Hardware GPU Scheduler was confirmed enabled.   09/25/2025 08:35:31 PM : Queue Started - Source File: C:\Users\XXXX\AppData\Local\Temp\Stampede Wagons PRORES v26.prproj - Output File: H:\YOUTUBE READY\TEST RTX5090 Stampede PRORES.mp4 - Preset Used: H265 Export  - Video: 3840x2160 (1.0), 60 fps, Rec. 709, 100 (63% HLG, 51% PQ), Hardware Encoding, Nvidia Codec, 00:38:04:51 - Audio: AAC, 320 kbps, 48 kHz, Stereo - Bitrate: VBR, 1 pass, Target 50.00 Mbps - Encoding Time: 01:00:53 09/25/2025 09:36:24 PM : File Successfully Encoded - Source File: C:\Users\XXXX\AppData\Local\Temp\Stampede Wagons H265_1_1.prproj - Output File: H:\YOUTUBE READY\1080p H265 to H264 Test HBIT Stampede Wagons H265.mp4 - Preset Used: - Video: 1920x1080 (1.0), 60 fps, Progressive, Rec. 709, 203 (75% HLG, 58% PQ), Hardware Encoding, Nvidia Codec, 00:38:05:03 - Audio: AAC, 320 kbps, 48 kHz, Stereo - Bitrate: VBR, 1 pass, Target 40.00 Mbps - Encoding Time: 00:24:38 09/25/2025 10:25:50 PM : File Successfully Encoded 09/25/2025 10:25:50 PM : Queue Stopped   09/25/2025 10:39:25 PM : Queue Started - Source File: C:\Users\XXX\AppData\Local\Temp\Stampede Wagons H265_1.prproj - Output File: H:\YOUTUBE READY\H265 to 264 Test Stampede Wagons H265_1.mp4 - Preset Used: Custom - Video: 1920x1080 (1.0), 60 fps, Rec. 709, 203 (75% HLG, 58% PQ), Hardware Encoding, Nvidia Codec, 00:38:05:03 - Audio: AAC, 320 kbps, 48 kHz, Stereo - Bitrate: VBR, 1 pass, Target 50.00 Mbps - Encoding Time: 00:20:55 09/25/2025 11:00:20 PM : File Successfully Encoded 09/25/2025 11:00:21 PM : Queue Stopped   I still have the non-beta software installed, I was going to try that also and report back. Not sure why I am not running into what appears to be a very consitent issue, if someone wants to share a project that fails, I would not mind trying it to see if I can render with Hardware. ... View more

there is zero impact to quality. Hardware GPU scheduling on allows the GPU to manage itself, which means your CPU is doing less work - in theory, this means you can get better performance in some cases.   Benchmarks however have shown mixed results, with same games show gains, others do not.   Its widely believed that this has a negligible effect on applications such as Adobe.   I have not tested it,  as my RTX 5090 is still in transit.   You can expect the same quality with the feature off or on - potentially a very minor performance impact, but there is no clear evidence for Adobe software.   If you are a gamer, you might want to turn it back on before playing games - but that requires a system restart. ... View more

its 100% confirmed Media encoder exports are takign 4 times longer than exporting the same clip with Premier directly....  I'll make a seperate post about this.     ... View more

@LUIS EDUARDO34647316dph8 I don't think there is a conspiracy here for the older cards,  sometimes software doesn't behave exactly correct and performance can change.  When you optimize for something, you can break prev work.  I suspect that's whats happened here, but I will have to dig around some more.  My 5090 hasn't arrived yet, seems to be stuck in shipping over the border, so hopefully next week I will it.   I will try and find some time to look at the 3090 performance over the weekend. ... View more

@pkirst There is another post where the 'known issues' for driver 581.29 still lists issues with export on premier.  Maybe that change log needs updating.  Previous posted by @robatst      ... View more

@robatst arg...  I forgot to check the known issues - good catch....  ... View more

@Elmasguapo can you please include the driver versions when posting - it makes it easier to compare ones installed version with what you seeing.   ... View more

there is also a new studio driver out today,  if the latest game driver corrected this, there is a good chance the studio driver has the included fix too....   There is a specific all out for memory leak fix when using nvenc hardware encoding;   Latest version is:       I have a Nvdia RTX 5090FE enroute, hope its here this week, and I should be able to jump into this testing pool as soon as it arrives.  Hopefully this latest driver actualy fixes it!  Not sure why I am going to suffer this pain but jumping in anyways.... ... View more

Is there any reason why we cant have the same quality of selection that is available in PhotoShop (beta that is, as the selection tool for ondevice in the PS Beta is really solid) for this tool?   ... View more

Maybe adobe can consider a new way to start a next,  like if you have a bunch of clips selected, maybe a icon shows up that you can click on to nest, without going into the menus/  and by extension, maybe nested clips can have a icon on them to 'un-nest' - just something to talk about....  maybe this idea sucks, I dont use nest too often, I tend to render and replace.... ... View more

@LUIS EDUARDO34647316dph8 the massive slow down in speed was a vibe I was getting also; but didnt have time to check it out... Do you have some numbers?  to me it feels like its about half the speed it was before.  But there is also a new media encoder beta out today so maybe it fixed that also.   As a side note, finaly, I have a RTX5090FE on its way to me,  so I'm about to dive into the pool of 5000 series issues next week I suspect....  hopefully I can confirm the typical performance returns to ME before I swap over to the 5090 or I will not be able to have soemthign to compare the peformance to. ... View more

Still not fixed in todays build of Media encoder.....     ... View more

@_nicdean fingers crossed for you and all of us!  This is a legit request...  BTW I tried to upvote, but upvotes are really hit and miss - if Adobe is looking for upvotes to prioritize these sorts of things, I think the inconsistent action of the voting button will make it hard to get a real sense of the interest in anything.   @Francis-Crossman Thank you for taking it up, and getting an project manager assigned to get this coded 🙂  I am looking forward to testing it. ... View more

This is not fixed in Sept 2 builds either, still broken.  I have a feeling its broken when you enable the new premier project file for AI object masks - but I cant go back and test this now.....   Versions: Media Encoder 25.6.050 Sept 2 Premier Pro 25.6.0.040 Sept 2   Versions are synced and error persists.     ------------------------------------------------------------ - Source File: C:\Users\XXX\AppData\Local\Temp\XXX 2025.prproj - Output File: H:\Ready For Plex\School XXX First Day Year 2 XXX Sep 1 2025\School UoT First Day Year 2 XXX Sep 1 2025.mp4 - Preset Used: H265 Export TK - Video: Based on source, Hardware Encoding, Nvidia Codec - Audio: AAC, 320 kbps, 48 kHz, Stereo - Bitrate: VBR, 1 pass, Target 50.00 Mbps - Encoding Time: 00:00:00 09/02/2025 06:30:57 PM : Encoding Failed ------------------------------------------------------------ Could not read from the source. Please check if it has moved or been deleted. ------------------------------------------------------------   LOG shows the following non helpfull information:   ... View more

@Stanley Arthur Its way too soon to present this to the public the way it is, and it seems you are also thinking that's the case.   For something that's just being shown now, its actually not bad, and has uses cases even in the current state.   What I think the issue is that the AI model they are are using to track/trace the object is too small.  I think they are building it for a 8gb GOU base hoping it works - because it seems there is very little GPU memory use when the process is selecting/tracking each frame.  They should consider offering a larger better trained model for those with more GPU memory, and explicitly tell people they need more memory.   One of the failing Adobe seems to have is trying to set really low hardware targets - and I think this is a mistake.   With that said,  I also think that this feature can work better than it currently is with even a 8GB AI model  -  I have provided some sample footage to them as I didn't think it was too complex for the AI to handle, but it was.  However, how it failed was not in how I would have expected it to - and at the same time, it would do other things like track a growing shadow coming into frame  from my selection that was rather small properly.   It was doing a good job of tracking a horse with a person on it, and not grabbing the shadow when I left it out (tracking separate).   I don't see any evidence that the AI model they are using uses any other tracked data to build a new track.  In mocha for example, I can track a easy part of the video frame, turn it off, and other tracks seem to do better because there is that one solid track already.  But Mocha is more for paper doll roto style, and honestly their ML track is also not great right now.   I am hopeful that Adobe puts the work into this and gets something very usable - it feels like its close with tweaking of the training models - with that said, and maybe I am wrong, Adobe needs to stop catering to 'non ideal' hardware and for these advanced features needs to call out exactly what they want/need - trying to get it to work on anything/everything is going to cause them reputation damage. (building for the lowest common thing is always bad news for the Professionals).   if Adobe wants I can test (and I am sure many others will) models up to 24gb GPU support right now (and I hope to get a 5090FE some day for even more memory) - I am convinced that a larger model and more varied training data will be helpful - but finding that training data will be difficult.     I think Runway had the luxury of starting it early and maybe (in my opnion) everything people uploaded was used to train their models, I'm not sure with the coverage Adobe got on AI model training they have had the courage to try something like that if only to train the tools - aka only train these tools and not for generative uses.   Overall, this tool shows promise, and its only in Beta,  I'm inclined to be open to where this might go - but of course they need to feel the pressure of clients and competition like to be be motivated to spend blood, sweat and treasure to develop it.  ... View more

@MyerPj thanks for the tip on holding ALT... I was going to attach a video, but didn'tt as I don'tt have a file share setup for it... hopefully my description makes sense.   Zooming to center works exactly as what I would expect...   I tried to show the zooming to mouse position, and it really doesn't seem logical for me,  it sorta goes towards to mouse pointer, but it moves the image around.  I think what I was expecting is that the zoom would be to exactly where the mouse printer is, and if your doing several zooms levels, it would still be to the same spot.   The way it works right now, if I center on say a person,  see about 21 second mark,  the zoom in doesn't go to the mouse position, and if I want to keep zooming into the same spot, I have to move the mouse pointer in odd ways.     What ever is exactly going on,  I was hoping to have it be able to continuously zoom into the mouse location as a target, without having to reposition every time I zoom a level.   This zoom method doesn't seem natural to me....   ... View more

I can confirm that the handtool (shortcut H) works.  What I would prefer is by default the handtool became active when your on the program monitor.  Right now, you have to change to selection mod (shortcut V) when you want to go back into the timeline to do something.   However, you can turn on scroll bars still, with the wrench icon on the bottom left of the program monitor then the menu selection below (this is from the beta version and I know the menus have been 'cleaned up' so non beta users may find it in a slightly different place.   I also don't like how zooming in, doesn't for me anyway, zoom into where the mouse pointer is.  If anyone knows how to deal with that please advise.       ... View more

Found something else;  trying to remove something that is selected in error; and while holding Alt to remove to it the following error comes up:   The part I am trying to remove from the mask is the little bit of grass in the top right corner of this picture.   Trying to add a missing section of this lamp post also generated the same error, and there was no way for me to add it to the mask to be included:   Trying to select this missing bit of the other lamp post always selects the windows and never the missing bit of lamp post:   These would all be a good use case for a brush tool I suspect where I can manually select the area.  It would link to the other detected masks, and use those to stay in sync if it was having a hard time tracking it (use relative planar data to stay locked on)   Another thing that is a workflow issue, is there needs to be markers for where the mask is drawn, and what part of the clip is tracked and not, and you should be able to use those to find your way back to the correct frame where the mask was created, and to skip to untracked portions of the clip.   Just found out the hard way by hitting the spacebar in error and it started playing the video, how hard it was to come back to where i was building the mask.     ... View more

@Arnold Nitin thanks for the feedback and response.  I have shared with you the sample footage with the horse and wagons so you may use it to test with also.   Apologies, I think I missed that helpx link in the original post - blew right past it to start testing I guess!  It was a helpful post, it confirmed a few things for me.  Also great to see that you expressly state your store the object data in its own file/folder and not inside the project file (thank you so much for making that decision, it will help with load times, and file corruption issues a great deal!).   As for the object selection tool, I think that needs to be expanded better or maybe I'm just thinking of how other tools worked too much and taking a second to switch my thinking.  allowing the object tool to detect other objects in whole might be something worth exploring, such as click on it, and tries to determine the object.  I will admit, this doesn't work great in another tool I am using to create garbage matte with, and it may turn out that using the pen to limit the auto detect object area is more ideal.  Could be worth looking at it closer, but its very usable with the pen, and circle selection tool modalities also.   I started using the rectangle tool, not the object selection, and then tried the pen tool - thinking that's what you meant - was about to write that didn't work - then after reading the the word "modalities" in the post, I tried the object selection tool again, but treated it like a rectangle, and wow!  it works as you described...  It took me a minute to understand that you could use the object tool, but treat it like a pen/rectangle....   not sure if there is a better way to denote that for users so we know it has those multiple ways of selection.   I used it to clean up and add to faulty masks - I am not enjoying the working in the program monitor - maybe you are thinking of a way to bring out the clip and allow more precise control etc , with a object mask workspace, or something - but working inside the program monitor is a little frustrating when you need detailed work - I hope you have some ideas around this.   there are a few other little bits such as not being able to move the viewport inside the program monitor while the tracking is happening to follow an object and zoom in and see the work its doing - it seems that when tracking starts all other processes are locked out.   It would be nice to see a stabilize feature to allow the program monitor stay locked on to the tracked object the whole time its processing also, and still retain the ability to zoom in and out while tracking.   I really hope you find a way to allow object track to be used a video stabilize feature,   I have not played with it, maybe that's already possible - but it can be very powerful to track and object and stabilize video against it (foreground and background).   I'm going to play with this somemore,  but so far its a very strong first (public) appearance of the feature. ... View more

So small bit of feedback,  beeing focused on the Beta these days, and they have a new feature in the public beta for masks... and guess what - they are storing the mask data externaly!  This is great news, and I hope they do the same thing with warp data also;     All I can say is Bravo - moves like this start to restore my faith in the devolopment - not only is this new mask tool really promising - but the very logical decision to have a folder for object masks and good documention around this a good step forward that I hoep they apply to warp stabilize now also.   ... View more