Video by @individualkex · Original TikTok post
- Views
- 704K
- Likes
- 87.7K
- Comments
- 693
- Shares
- 5.3K
- Saves
- 20.2K
- Engagement
- 13.3%
- Top 18% of 17 DeepSeek posts by views
- 1.1K views a day since it went up
- 12 of 17 use this format (App screen explainer)
- @individualkex has 1 video in the library
Creative DNA
- Format
- App screen explainer
- Shot as
- Talking head · Human creator
- Hook
- Curiosity gap · spoken
- Tone
- Informative, Objective, Fast-paced
- R1
- AI
- Open Source
- Janus-Pro
- LLM
- Tech
DeepSeek R1. What is it?
The recipe
CopyPro- First line
“DeepSeek R1. What is it?”
Curiosity gap · spoken · 1.0s long
- Beats
- 7 beats · 0:00–0:51
- First shot
- The text DeepSeek R1 appears on a black screen.
- How it's shot
- App screen explainer · Talking head · Human creator
Where: Home office or studio
- Made for
- Tech enthusiasts, AI users, and developers interested in open-source tools.
Mood: Informative, Objective, Fast-paced
All the notes
Shoot script · Remake brief · AI remake setup · Why people watch · The pitch · Script · Caption
Make your version
Shoot script
Remake brief
The same idea for your app, beat by beat.
The model we would use today
Source notes
Why people watch
- The video addresses a trending topic immediately, which captures the attention of people already searching for information on the brand.
- The creator breaks down complex technical concepts like reasoning versus prediction into simple, understandable terms.
The pitch
The pain
Proprietary AI models are expensive, closed, and slow.
The fix
DeepSeek R1 is a better, faster, cheaper, and open-source alternative.
Why you believe it
The creator explains the technical difference in training methods, showing that R1 reasons instead of just predicting, and highlights the new Janus-Pro model.
The creator explains the rise of the DeepSeek R1 AI model, contrasting it with proprietary models like ChatGPT.
Script
EN · speech and on-screen text
Transcript
- DeepSeek R1.
- What is it?
- A language model like ChatGPT, but better, faster, cheaper, and open source.
- It also happens to be a small side project.
- For a Chinese quant trading company, many are unhappy about it.
- But in reality, it's not US versus China.
- It's proprietary versus open source.
- And open source just jumped way ahead.
- So how does it work?
- Normally, language models are shown a bunch of examples of questions and answers.
- These are trained to predict the answer.
- DeepSeek R1 doesn't do that.
- Instead, it's given the question, but no answer.
- It has to come up with its own answer, which is then graded.
- So it's forced to reason instead of just predict.
- So how do I use it?
- Look up DeepSeek R1 and use it.
- That's helpful.
- There's also an app, and since it's open source,
- you can use it anywhere else, like HuggingChat, which also has an app, or your own computer.
- I made a tutorial on that.
- So what does this mean for breaking news?
- DeepSeek just released another model, Janus Pro.
- Another side project.
- It can generate images better than DALL-E or Midjourney.
- You can try it now on Hugging Face.
- So what does this all mean?
- AI just got a lot better, cheaper, and more transparent, which is good news for almost everyone.
- Thank you for watching.
On-screen text
- deepseek R1
- what is it?
- better! faster! cheaper!
- but in reality
- proprietary vs open-source
- normally LLMs
- Q: how many R's are in the word strawberry? A: 3
- i made a tutorial
Caption
DeepSeek situation is crazy
Sound: original sound - individualkex · Region: US · Stats saved Nov 4, 2026
Top DeepSeek videos
Brand strategyApp screen explainer





