Help:Creating Images and Videos

Help page
Revision as of 18:05, 25 January 2023 by PrivateTarkus (talk | contribs) (PrivateTarkus moved page Help:Creating Images to Help:Creating Images and Videos)

Video Content

Video content is primarily used in Combo Theory boxes and other places on the site that benefit from visual information.

When to use youtube to host videos:

  • Use youtube for anything over 30 seconds
  • Use youtube if you want high quality video

When to use locally hosted videos:

  • Upload directly to Dustloop if you have small, short videos focused on a single concept
  • Compress videos uploaded to dustloop with the information in the next section

Locally Hosted Videos

In order to keep file sizes low, it is strongly advised to compress them. Removing or lowering the sound of the videos is also recommended as they will be played at maximum volume by default. The following method is recommended:

Download ffmpeg

Windows: https://ffmpeg.org/download.html
Linux: sudo apt install ffmpeg

Go to ffmpeg

Open a cmd/terminal and type the following command:

For windows:
cd [directory where you have ffmpeg]
Example: cd C:\Users\Dustloop\Downloads\ffmpeg-master-latest-win64-gpl\bin
For Linux:
This step isn't required. However, it might be necessary to use "usr/bin/ffmpeg" in the next step instead. If it still doesn't work type "whereis ffmpeg" in a terminal and use that result instead.

Convert the video(s)

In the following section, "video_name.format" refers to the name of the video to convert and its format (mp4.webm) and "output.webm" the name the compressed video should have. To make it easier, move the video(s) to the same directory as ffmpeg.

For 16:9 games (Guilty Gear, DNF, Blazblue...)
ffmpeg -i "video_name.format" -vf scale=-1:480 -b:v 1000k -vcodec libvpx-vp9 -threads 4 -y -an "output.webm"

For 4:3 games (+R, Melty, most doujin games)
ffmpeg -i "video_name.format" -vf "crop=iw-480,scale=-1:400" -b:v 1000k -vcodec libvpx-vp9 -threads 4 -y -an "output.webm"

Those commands divide the size of the videos by a factor of 8~9.

Details:

  • vf scale=-1:480: Resizes the video to be in 480p.
  • b:v 1000k: Reduces the video bitrate.
  • y: Overwrite the video.
  • threads: Decrease the amount of time required to compress the video but increases CPU load.
  • an: Removes the sound.

Trim the video:
ffmpeg can also be used to trim part of video with the -ss and -to options. They are respectively used to define the start and stop times.
For example:
ffmpeg -i "video_name.format" -ss 0:03.000 -to 0:08.000 -vf scale=-1:480 -b:v 1000k -vcodec libvpx-vp9 -threads 4 -y -an "output.webm"
Only keeps the part of the video between the 3rd and 8th seconds.

Keeping Audio:
If sound is required for audio cues, it is recommended to lower the volume of the video as they will be played at maximum volume by default. This can be done by replacing -an with -af "volume=0.1"
ffmpeg -i "video_name.format" -vf scale=-1:480 -b:v 1000k -vcodec libvpx-vp9 -threads 4 -y -af "volume=0.1" "output.webm"

Other parameters tested:

  • fps: Had very little impact on video size.
  • crf (constant rate factor): Had very little impact on video size. Adding --crf 40 can help reduce the video size a little bit but will affect quality. Increasing the value lowers the quality/size even further.

The two primary goals for the images are

  1. Make the attack easily identifiable by players who only slightly know the character and have seen each attack once or twice at most.
  2. Be a visual aid to explain how an attack works.

Image Guidelines

  • Use sprite rips with a transparent background for images even if the sprite does not use all the special effects used in game. As long as the attack is still easily identifiable as that attack, it's fine.
  • If no rips are available, use high quality PNG screenshots taken with a capture card or straight from the PC version of the game.
  • Characters should use default colors and face right (as if they are player 1)
  • Name images with the following format: GameAbbreviation CharacterFullName MoveNumpadInput.
    • Examples: Sol Badguy's 5P in Guilty Gear -Strive- would be named GGST Sol Badguy 5P.png. His Gun Flame special move would be named GGST Sol Badguy 236P.png
    • For moves using multiple images, add the image numbers after the input. For example, Ky Kiske's two-part Stun Dipper in Guilty Gear -Strive- would have two images named GGST Ky Kiske 236K 1.png and GGST Ky Kiske 236K 2.png
    • For moves with multiple possible button inputs resulting in different move versions, substitute X e.g. GGST Sol Badguy 623X.png
  • Hitboxes should be named similarly, with "Hitbox" at the end.
    • Examples: Sol Badguy's 5P in Guilty Gear -Strive- would be named GGST Sol Badguy 5P Hitbox.png. His Gun Flame special move hitbox would be named GGST Sol Badguy 236P Hitbox.png
    • For moves with multiple hitboxes, the number should come before "Hitbox". For example, Sol's 5K hitboxes would be named GGST Sol Badguy 5K 1 Hitbox.png and GGST Sol Badguy 5K 2 Hitbox.png
    • Only use letters and numbers, -, _, (), and ~. For example, Sol's j.P should be named GGST Sol Badguy jP.png, Leo's [2]8X should be called GGST Leo Whitefang 28X.png, and Dragon Knight's charged 5S should be called DNFD Dragon Knight 5S-Charged.png
  • Nearly all images that have text use the font Arial bolded. Please continue to use it for the sake of consistency.
  • Combine multiple sprites together if the attack does an animation that can not easily be understood by one sprite alone.
  • If the move in question is a grab or command grab, make sure to include the victim in the image. As well, never upload just the whiff animation of these moves by themselves. Always accompany them with a successful grab.

Image Editing Tools

  • Photoshop
  • Gimp
  • Paint.NET
  • Krita

Consider taking a learning course on image manipulation basics to help you learn what tools are at your disposal in these programs.

Taking Screenshots

  • Have the character in default colors/costumes
  • Make sure the stage doesn't obscure the attack. Find a stage (or part of a stage) with colors that don't blend into attack
    • Also consider taking screenshots while other effects darken the stage like RCs in Guilty Gear Xrd or Overdrive in Blazblue CF
  • Be consistent with the stage. Use one stage for all the screenshots for that character. Using a different stage for another character is fine.
  • Hide things that aren't related to the move in question if you can.
    • Ex: In Guilty Gear XRD, Many of Ramlethal's normals don't involve her giant swords, so don't include them (or crop them out) if you can
    • Ex: Granblue Fanatsy Versus, Zooey can have her dragon with her, don't include it unless it's part of the attack
  • Avoid having the opponent in the screenshot unless needed (Ex throws, proximity normals, etc.)
    • If they are needed, make sure they are easily distinguishable from the attacker (no mirror match, different colors)
    • It's okay to crop out parts of the opponent like an outstretched arm
  • Try to avoid having the HUD in the screenshot if you can. You can move HUD elements up/down slightly in the game's display options. Some games even allow you to hide the HUD completely (or via mods).

Images Need to be Readable When Zoomed Out

Remember, these images are shrunk on character pages, so make sure text and diagrams are still readable when shrunk. Before uploading an image, open it on your computer and zoom out to see if the image is still readable.

  • Crop the image so that extraneous details such as long ponytails do not take up too much space.
  • Consider moving objects closer together - these images are to help readers identify the attack and not necessarily show the attack's range. For example if the attack hits far away, such as Bedman 3H, or Cell's command throw, it's okay to move the hit effect closer.
  • Use colors for text and arrows that don't blend into the background. Consider adding an outline to them to better help them stand out

Images with Transparent Effects

Building images with transparent effects are crucial for helping with move clarity and readability. This is especially true if a move is effect-heavy. User Wakeup720 has a small guide to follow for helping to build transparent effects:

  • Get a black screen mod for the game in question. These are usually found on modding sites like Gamebanana or ModNexus.
  • Set your game resolution to 1280x768, keep resolution on.
  • Perform the move and use a freeze-frame technique to get a good image of the move. On Steam, you can do this with F12. Most fighting games allow you to hide the U.I. in training mode by pausing and pressing select.
  • Get a snip of the move that includes all the effects and open it in Gimp.
  • Get the character sprite from a spriters' resource and overlay it on top of the image in Gimp. Blazblue has [Boxdox], but you can also take a screenshot twice and use that.
  • Use Select By Color to select all of that sprite, then go to the image layer and cut/copy and paste that selection into a new layer.
  • Go back to the image layer and select Color > Color to Alpha, and set the color to black.
  • Then hide the sprite layer, and tadah! Transparent image.

[Graphics Frame Analyzer] can also be used to eliminate draw calls in an image and isolate specific images, but that requires a bit more advanced learning.

Use Multiple Images to "Cheat"

Since these are still images trying to show what an animation looks like, it's necessary to cheat a bit to create images that give the illusion of the whole attack.

Think of it like adapting a scene in an action movie into a comic panel; the same scene in a comic panel will need to combine multiple images and cheat time/space to give the same impression.

  • Ex: the image for Vegeta's Consecutive Energy Blast does not exist in the game, it was made by putting multiple blasts together to give the visual illusion that he has quickly shot multiple shots
  • Ex: Millia's Chroming Rose does not spawn the roses that close to her. The image was made by running forward, then backdashing and running forward again.
  • Ex: The sprite of Hilda is pasted near her attack. In the game, she recovers too fast for her to be in the pose when the attack activates. We do this since her pose and the attack are the two most visually recognizable parts of the attack, even though they appear at different times.
  • Ex: The Metera screenshot is 2 images combined for the same reason as the Hilda example

Screenshots With Invisible Backgrounds

See here for a Video Explaining Hitbox images are created in Guilty Gear Strive.

Back in the past for Guilty_Gear_Xrd_REV_2, it was primarily the work of one dedicated user: User:Johhny. He wrote up a guide on how he does it here.

Exportation and Upload

Once the image has been created, you can export it to a PNG and upload it to the site. We recommend using PNG-8 as the file type for image uploads, since the compression helps keep file sizes low without losing much quality.

After all that is done, navigate to our Special:Upload page (which can be found under the "Tools" heading of our sidebar) and upload the file with the correct title. The naming convention of Dustloop images is "(Game Title Abbreviation) (Character Name) (Move Input or Move Name)". Choosing between move name and move input doesn't matter much when uploading a file, so long as the image itself is correctly titled. You can click on the images at the top of this page to see this naming convention in action.

Renaming Files

If you need to rename a file due to a bad name, use the Move command.

  • Navigate to a file (ex: File:BBCS_Ragna_the_Bloodedge_5A.png)
  • At the top right of the page, click ... > Move
  • Follow instructions on the page to rename
    • Uncheck the "Leave a redirect behind" option. A Redirect makes it so the original file name still exists, but it points to the new file name. For the scenario of simply renaming a bad file name, this is not needed.
click "...", then click "Move"