Blender VSE improvements: Road to VSE 2.0
Veteran Blender developer Sergey Sharybin has started work on the video sequence editor as part of the VSE 2.0 project. Its main goals are to improve the usability of the VSE and drastically improve its performance, making it a viable option for the complete process of editing a movie project. The work plan is divided into four milestones reflecting these goals, starting with playback performance and reliability.
Within a week of Sergey starting to code, the VSE already boasts a 24% performance increase in some cases. Sergey also implemented a “zoom to fit” feature where the video preview automatically resizes with the window, making it the default behavior. The old behavior of the video preview having an independent scale, which is more reminiscent of photo-editing programs, is still available. Twitter user @tintwotin demonstrates the feature:
"Zoom to fit" in the Video Editor Preview by #sergeyontwitter #b3d #vse pic.twitter.com/eJZ9woPfCQ
— tintwotin (@tintwotin) August 14, 2020
Another annoying thing is the annotations, not rescaling when resizing the output resolution… Please Sergei, help!
Do you mean the texts of Text Effect strips?
While I appreciate the effort, there are many free and open source video editors available. It seems to me that the Blender Foundation’s limited resources would be better spent on developing basic 3D related VFX features (for example, shadow catcher functionality for Eevee).
I believe it is not about having open source vse, its about expanding blender to heights unimaginable for most other applications. And to be honest, I’d rather see redundant features polished to the max in blender than having to put up with the mediocricy of a dozen other software. All the more things to learn and forget. Blender is supposed to be the go to one and only software by the looks of the current direction for all things indie animation and vfx. And I fully support it. I would love to see a krita-like painting feature integrated into blender, and even an audacity-like audio editing. Because everything the Blender team puts out turns out pretty good, and having things integrated is a good choice in my view.
Just look at the adobe cc pack. They have various applications that can work together to a certain extent. If Blender could surpass the capability, it would seriously disrupt the market.
Blender devs, please ignore the above comment. ;)
(the first comment, not the other reply)
There are plenty of other areas in Blender where this would apply too, like text editing and texture painting. Blender strives to be a comprehensive platform for your entire workflow, so I think it makes a lot of sense to have an integrated VSE. Its close connection to the renderer and compositing make it very powerful and offer a very flexible workflow.
The same thing could be said about EVEE. Why develop a real-time rendering engine when the game engine it would have been mostly useful for has been canned. Ogre is a free and open source real-time graphics renderer, why not integrated it Blender to make Archivz and such. Will EVEE play catch-up to the plethora of real-time render engines out there and then be dump because it will have become a burden and irrelevant like the game engine? Cycles needs so much more improvement that independent developers sell “better cycles”. So I think the VSE will also be a waste of time and eventually be abandoned unfortunately. Blender uncompressed is now several hundred megabytes when it used to boast its light weight. Blender is now becoming an average tool with paid-for addons to compensate for its “jack of all trades master of none” direction.
I do all my video editing in VSE and have done so for at least the last 5 years.
Me, too! The VSE has its weaknesses, but I like it very much.
I do all my accounting on a piece of paper but I’m not a million dollar company.
More than 10 years ago I was interested in doing a google summer of code to improve the VSE.
I was pretty much told: It’s good enough for the Blender Open Movies, we don’t care about anything past that.
At the time there was lot’s of enthusiasm for big moonshot projects with low chance of success. A project with lots of small incremental and easy to achieve goals didn’t get any interest.
I was so discouraged, I never submitted a proposal or took another look at the Blender code.
I’m glad it is finally a priority.
If you still have an interest in contributing: https://blender.chat/channel/vse
Request blender to include denoiser for video editing to improve video quality.
Glad to see the VSE is getting some attention (again).
Here’s hoping for a lot of love to the VSE. It’s a VITAL part of Blender!
Blender Video Editor essential add-ons, which are not included in Blender:
Parenting, volume meter, tagging etc.: https://github.com/snuq/VSEQF
Transform tool add-on: https://github.com/doakey3/VSE_Transform_Tools
Generate proxies: https://github.com/cgvirus/blender-vse-easy-proxy
Parallel Render: https://github.com/elmopl/ktba/blob/master/scripts/addons/parallel_render.py
Additional 125+ add-ons for the Blender Video Sequence Editor:
https://blenderartists.org/t/video-sequence-editor-news-add-ons/1188770
Holy cow, 125 just for the VSE?! That’s amazing.
Well, my opinion may not be very popular, but I have an impression that the VSE is a part of blender kind of glued to it, and does not have any deeper reason to have it there. Due to that I feel like it would make more sense to redirect efforts more towards other parts of 3d workflow/featureset, and leave VSE behind, like it was with BGE at some point. Here is the reasoning behind statement above:
– non-linear video editing included in the package is not actually synergic with rest of Blender. The thin layer of connection between what Blender really is and VSE is ability to render directly into VSE – in practice no one is doing this for any more serious work than rendering test cube. then, If we assume we are not rendering 3d view directly to VSE, we are getting just a separated non-linear video editor, being glued by force into blender application and using blend files to store data. No clear benefits of having it there, but a lot of challanges connected with a need to fit it into blender structures, UI and UX.
– current VSE misses lots of helpful tools and has really bad UX (and I am saying this as an experienced Blender user). If we add to this performance issues, its simply monstrous amount of work to make it up to speed with competing video editors
– competition is really crowdy, with superb tools, that are free to use for non-commercial projects, or really affordable for commercial ones, as well as vast choice of open source picks (just to mention HitFilm, DaVinci Resolve, Lightworks, ShotCut, OpenShot and more)
With that said I just don’t feel like chasing the competition from the place where VSE is makes any sense, expecially if we are talking about it in a context of development of Blender, which is primarily a 3d creation suite. I must admit I have limited experience with VSE, but the reason behind that is actually what I wrote above – other software is simply making much better job in that area, plus there is literally no benefit of keeping video montage in blend file. With my humble experience in that area, I simply don’t feel this effort will bring a lot of value into Blender.
I may be wrong of course, just sharing thoughts.
The VSE needs to be better integrated with the rest of Blender, so dynamic unrendered workflows will be possible all the way from having a vague idea to final render. This way Blender will be a tool for storytellers it was meant to be.
Take a look at this example:
“I’m super happy that they’re not removing the Video Sequence Editor, because I think that would have been a huge mistake” – Rob Silvestri, Tangent Animation: https://youtu.be/IXplyBLw21A?t=623
In the above example the VSE is used for editing unrendered animatic/grease pencil shots which is possible through scene strips. This allows for a quick dynamic workflow where you’re able to “shoot” footage while “editing” it: https://youtube.com/watch?v=AG52ytTa_JU
Better integration would also mean that the contents of the Sequencer would be a data-block you could assign(unrendered) to ex. an input node(for VFX/grading etc.), the background of the 3d view(for ex. gp drawing on on top af edited video), a texture etc.
When these kinds workflows are possible, there will be no more discussion of competing with existing nles, because none of those would be able to handle unrendered material inside Blender like Blender could.
The reason for the current slow performance and the limited workflow is that the VSE runs on CPU and the 3d view on GPU. In order to make the VSE run on GPU, a refactor of the VSE code is needed. The VSE maintainer is working on this currently. (He just made transforming in vse 3,5% faster by multi-threadning it)
The rendered video handling should be secondary to the unrendered, but it seems like speedups, like the ones Sergey found, which changes two lines of code, but gives the used ffmpeg lib a speed up of 24%, are possible. On top of this there are a lot of new ffmpeg hardware accelerated options for ex. Nvidia cards which can speed up video handling even further, which currently aren’t enabled(and maybe still a bit rough around the edges in ffmpeg?). Colorhandling also seems to be an area where improvements can be found and bugs can be fixed, but a major update of the lib handling that seems to be close to a release.
These things will be and should be addressed first, then the tool set can be improved.
Currently great add-ons like Power Sequencer(comes with Blender) and VSE Quick Functions can be used to bypass some of the toolset shortcomings. The above mentioned code refactor will also pave the way for better coding of built-in tools in python, which means that a lot of tool code can be lifted into Blender from the add-ons, this will also allow for the community to contribute with a fast development of the editing toolset.
All of this is of course a lot of work and needs more devs to implement, but when it is done, Blender will truly be a tool for storytellers.
Nice examples of workflow you have shown, I was not aware of those and it looks really interesting. Still it’s not justifying for me the efforts on expanding VSE. You see, those are very specific uses of blender, which are utilizing to the max the fact that VSE is tightly integrated as a part of it (so – point taken it makes sense to have VSE there for some cases). Still it’s a fraction of how ppl are ustilizing video sequence editors, and what’s more important – Blender VSE development seems to be oriented to improve actually “the rest” of use cases, no particularly those you have pointed.
So then the question, if benefits of pumping up VSE justifies effort and cost, is still valid for me. There will be always ppl using VSE, same as ppl were using Blender internal till the last hour, and same with BGE. Still, due to the particular blender needs, competition and available tools sometimes it makes more sense to let sth go, than fight for it at all cost.
Maybe I am too much focused on what I am usually doing, but still I have strong doubts about pushing devs to work on VSE now, when there is so many topics that would push Blender to another level in it’s core scope.
Again, just sharing thoughts, maybe there is strong vision in Blender Institute, where VSE is a must to complete the picture.
I use the VSE all the time – this is good news!
I have made many video. tutorials for Blender. More than 200 think. Almost of them have been made using Blender. I glad to hear that the Blenders video producing capabilities are planned to be expanded.