Beluga videos look like they were made by taking screenshots of a Discord server and stitching them together. That is how the early ones were made, and it is why most tutorials still tell you to open Discord, type the conversation with alt accounts, screenshot every message, and line them up in CapCut.
You can do that. It takes about forty minutes for a twenty-second video, and the timing comes out wrong on the first three tries. Here is the faster way.
What you are actually making
Strip a Beluga video down and there are four parts per message:
- A wait before the message appears.
- The message pops in. A ding plays.
- The message holds on screen for a beat.
- Sometimes a punch: a boom, a zoom, a reaction emoji landing late.
The joke is almost never in the text. It is in how long the screen stays quiet before the reply. A three-second hold before "yes" is funnier than any line you could write. Once you see the format as timing rather than screenshots, the whole process changes.
Step 1: write the chat
Open dingcut.xyz/new. No account needed for this part.
Add your characters. The classic setup is three: a straight man, a chaos agent, and one who only shows up for the punchline. Give each one a name and an avatar. Then type the conversation the way you would type it in a real chat. Enter sends the line. Tab switches speaker.
Keep lines short. Beluga lines are five words or fewer most of the time. "doesnt exist yet" reads faster than "The video doesn't exist yet." Lowercase, no period, no apologising.
Step 2: pick a rhythm
Dingcut has three rhythm presets. They set the wait and hold for every message so you do not have to time anything by hand.
| Preset | Wait | Hold | Use it for |
|---|---|---|---|
| Slow burn | 0.5s | 1.6s | Drama, dread, a story that builds |
| Beluga | 0.3s | 1.1s | The classic pace. Start here. |
| Machine gun | 0.1s | 0.5s | Arguments, spam, "ok ok ok ok" |
The numbers came from timing real Beluga uploads with a stopwatch. Beluga is the right default for almost everything.
Step 3: break the rhythm where the joke lives
This is the part screenshot tools cannot do. Click a message and the inspector on the right shows its wait, hold, sound, and punch. Change one line.
The pattern that works most often: extend the hold on the question, then boom on the answer.
- "are we in it right now" gets a 3 second hold with crickets under it.
- "yes" gets a boom and a zoom punch.
Two overrides. Everything else stays on the preset.
Step 4: sounds
The ding is the Discord notification sound and every Beluga video is built on it. Dingcut ships 49 original sounds, all synthesized in-house so you can use them commercially. Ding, vine boom, air horn, join, leave, incoming call, crickets, sad trombone, and a few dozen more.
Each message carries its own sound. You can also upload your own. If you have a specific meme sound you need, drop the file on the message.
Step 5: pick a skin
Discord is the original Beluga look. The same script re-exports as iMessage, WhatsApp, Telegram or Messenger with one click. This matters more than it sounds. A Discord story and an iMessage story from the same script are two uploads, and the iMessage version tends to do better on TikTok.
Step 6: export
Press Export. Dingcut renders the MP4 in your browser using WebCodecs. There is no upload and no render queue, so a twenty-second video takes a few seconds. You get a 1080x1920 H.264 file ready for Shorts, TikTok or Reels.
Free accounts get three exports of up to twenty seconds with a small watermark. Pro is $2.99 a month for unlimited exports up to ten minutes, no watermark.
Common mistakes
Too many messages. A good Beluga short has eight to fifteen lines. If you are past twenty, you have two videos.
Every line gets a boom. Then none of them land. One boom per video, maybe two.
No silence. Beginners fill every gap. The pause is the joke. Give the setup room.
The punchline is too long. "yes" beats "yes we are literally inside the video right now." Cut until it hurts.
The three-minute version
Write ten lines. Leave the preset on Beluga. Pick one line for a long hold and one for a boom. Export. That is the whole workflow, and it is under three minutes once you have done it twice.