Best Practices: Making Videos

This document is a living guide to help us improve the quality and consistency across all our tutorials. It focuses on practical tips about writing, recording, and editing. For notes on education, see Education in online tutorials.

Creating great videos starts with planning, writing, and recording good footage. I detail my approach to this in the video below.

In short: check-list

To record and edit efficiently:

Record multiple videos in a row, ideally entire series at once. For example, one chapter in a course.
Take notes along the way. Write down:
- Important edits you do not want to forget like an entire section that needs to be cut out. This saves you time later on.
- Write questions and answers for quizzes, topics for complementary lessons, ideas, etc.
Organize your files as you go.
- Name good recordings.
- Delete bad recordings.

When recording:

Prepare your workspace.
- Review your notes and the list of videos to record.
- Before recording each video, review the changes or features you have to cover, or your demo project.
- Open all the programs you need: Key Mon to screencast keys, Open Joystick Display for the game.
Make the main program you’re recording full-screen.
Warm-up your voice.
Do a short test recording to ensure that everything is working as expected.

When it comes to editing:

Use the auto-import feature of Power Sequencer.
Normalize audio tracks. To do so, listen to the audio and compare it to a reference video or audio track. You cannot rely on the waveforms there; their amplitude isn’t representative of the perceived volume.

Video recording

You will save a lot of time if your source material is clean and your speech precise. If you don’t take the time to plan, you will limit the quality of your tutorial.

You should always record:

At 1920x1080, in full-HD
At 29.97 Frames Per Second, also called NTSC or NTSC film

Your footage should be smooth, not choppy. Be careful that the framerate does not drop in your recordings. To do so, I recommend you to do a test recording at the start of recording sessions or when you change your recording environment. This way you’ll ensure that the visual quality and audio levels are working as expected.

You should suppress or close any program that can distract the viewer from the tutorial:

Turn off Notifications
Close chat applications
Make the programs full-screen
Suppress system sounds

Avoid waving the mouse cursor as you talk. Put the mouse or stylus down when you’re talking, and you’re not pointing at anything in particular. The motion can be distracting.

Set the desktop audio channel to record at a low volume so it doesn’t cover your voice. I keep it at -20db to -30db most of the time in OBS Studio.

OBS mixer

When you make a small mistake or need to re-record a sentence, clap in front of the mic. This produces a sharp vertical line on the audio waveform that makes it easy for the editor to find and remove the faulty parts.

Clap on audio waveform

Setting up the microphone

The closer you are to the microphone when you record, the louder your voice will be relative to the sound reflected by your environment. You should place the microphone no more than 50cm away from your mouth if you can.

Don’t place your mouth right next to the microphone if you still move your head while recording: your movement will cause significant changes in your voice’s volume.

The voice should be normalized at around -6db. Use other YouTube videos or one of your recordings with good audio levels for reference. Always check your edit against the reference before exporting your video.

Preventing mouth noises

Mouth noises happen when your mouth is dry, or after drinking coffee, tea, and some other beverages.

If you drank one of these, try to brush your teeth, tongue, and palate before recording.

Keep a bottle or a glass of water next to you to stay hydrated.

Warming up

The mouth and vocal folds move thanks to muscles. To articulate to your fullest and to get the most out of your voice, you need to warm up, like before any exercise. At least for a minute or two.

The longer the recording session, the longer you should warm up.

Here’s a short vocal warm up from Julian Treasure’s TED talk. You want to warm up your lips, your tongue, and your jaw. The siren exercise then focuses on the vocal cords. To improve pronunciation of specific sounds, add a few tongue twisters: they’ll serve both as an exercise and complete your warm up.

Take your time

If you’re struggling to explain something while recording, clap, and take the time to breathe and to think. Rehearse the complex idea in your mind or in front of the microphone. You can leave the recording running in the meantime. When you’re ready, do a clean and confident take. It’s easy to edit the long silence out later.

Program layout

Use one of the default interfaces and, when possible, the default interaction mode for the program covered in the video.

E.g. for Blender 2.80 and up, left click is the default to select and grab objects. As viewers stumble upon videos organically and generally won’t watch them in a particular order, the tutorials should stay close to the programs’ default settings. There are some exceptions for tools like the shell, vim, or emacs that, once you’ve learned the basics, you are meant to customize a lot.

Make the program full-screen whenever possible. Many programs use the F11 key or a related shortcut for that.

For Godot, before recording, go to Editor -> Editor Layout -> Default to reset the interface.
For Krita, I recommend using the Big Paint workspace over the default, as it centers the canvas on screen. It also shows the Layers, color palette, tool options, document overview, brushes, toolbox, and advanced color selector: everything you need to paint is visible on screen.

Use a program to screencast key presses

Use a program like Key-Mon to display key presses, mouse clicks, etc.

Screenshot of key-mon

You can install Key-Mon using the Python package manager: pip install key-mon.

That way, even if you forget to mention a keyboard shortcut, the viewer can see it in the video. This also reinforces vocal cues.

Video editing

Avoid abrupt cuts in both the audio and the video. You should cut mistakes and long silences or hesitations, but avoid cutting too much between words.

When editing live action footage, like interviews that are face camera, inserting many cuts can make the video feel robotic or sped up. There’s a natural rhythm to the human speech we should preserve.

Make the flow of the video easy to follow

Assume the viewers are facing new information and have a big cognitive load when watching educational videos. Anything that makes it harder to keep track of what the teacher is doing increases that load:

The lack of visual continuity between shots, like jumping between files without mentioning it.
Making changes in the project without communicating them to the student.

That is why you should use the visuals, audio cues, and your voice to make understanding minor details of the tutorial as comfortable as possible.

Mention and show things such as:

Navigating through the program.
Navigating through code.
Or when you bring up a topic that’s not illustrated by the original recording.

When navigating around the interface or documents, show how you get there with the mouse cursor. If possible, also tell the viewer where you are going.

Use the animated arrow to point at UI elements

Often, there are situations where we have to move from one side of a program to another.

For example, after creating a node in Godot, in the docker on the left, we configure it in the inspector on the opposite side of the screen.

The blender video.blend template includes an animated pointer arrow.

Use this arrow to help guide the student’s eye to small icons or when jumping to a distant area of the interface. Especially when it’s not obvious on screen.

Blender video editing template

Use our video editing template to get your projects started. This requires some assets, that you need to place in ~/Videos/assets.

If you use the Gnome desktop, you can place it in the ~/Templates directory to register it as a template when creating new files.

The template comes with three layouts:

Video Editing, to cut and sequence strips.
Color Grading to tweak the colors after editing.
Rendering, to change render settings and preview the render.

The video editing layout displays the graph editor in the top-left to animate strips and transitions.

Using the animated arrow

You can use the animated arrow to point at specific elements, for example, at icons in a Godot tutorial.

To do so, you’ll need to install and activate the VSE Transform Tools add-on.

I placed the arrow before frame 0 on the editing board so it won’t appear in the render. To add a new arrow and position it:

Select all three strips that make the arrow
Press Shift D, or D with Power Sequencer, to duplicate it.
Select the topmost Transform strip.
Place your mouse cursor over the video preview, and press G, R, or S to move the arrow.

The first transform strip has some animation on the translation and blend_alpha channels. The topmost transform effect allows you to move or rotate the arrow while preserving the animation.

Automating the boring parts

You can save time on repetitive tasks and increase your productivity following conventions and using simple programs.

Using conventions

To be as efficient as possible, follow the same process and naming conventions to record videos. This helps to later batch-process videos using code.

Try to record several videos in a single recording sessions and name them as you go. For series, start each recording with a two-digit number. For new videos, change the number.

Here is an example with three video projects recorded in five segments:

01.intro-fsm.flv
02.importing-start-dir.flv
03.state-class-01.flv
03.state-class-02-fix-end.flv
03.state-class-03-fix-end-2.flv

When a video is made of several recordings, you can number the parts to keep them sorted.

Create a folder for each series where you will place all the recordings.

With these naming conventions, you can then create projects in batches using a command line script.

Creating video projects in batches

The fish shell script below creates a blender video project for each set set of recordings, following the naming conventions above.

It creates a directory for each video project, create a blender project from a template, and move all the footage to a subdirectory called “footage” in each folder.

For example, from the files above, the script will create the following structure:

.
├── 01.intro-fsm
│   ├── 01.intro-fsm.blend
│   └── footage
│       ├── 01.intro-fsm.flv
├── 02.importing-start-dir
│   ├── 02.importing-start-dir.blend
│   └── footage
│       ├── 02.importing-start-dir.flv
├── 03.state-class-1
│   ├── 03.state-class-1.blend
│   ├── footage
│   │   ├── 03.state-class-1.flv
│   │   ├── 03.state-class-2-fix-end.flv
│   │   ├── 03.state-class-3-fix-end-2.flv

To use the script, you need:

The fish shell installed.
A template blend file called video.blend in your ~/Templates/ directory.

How to use:

Copy the code below.
Open the folder that contains all the videos with the fish shell.
(Optional) Make a backup of the files.
Paste the code below in your terminal.

#! /usr/bin/fish
set temp_file mktemp
for v in *.flv; string replace .flv "" $v >> $temp_file; end
for i in (cat $temp_file | uniq -w 3)
    mkdir -p $i/footage
    mv (string sub -l 2 $i)*.flv $i/footage
    cp ~/Templates/video.blend $i/$i.blend
end
rm $temp_file

The script creates the blender projects for you. Then, you can:

Run bpsproxy to generate all proxies.
Use a command like for i in */*.blend; blender $i; end to open each video and edit one after the other automatically.