TikTok Series Plan: Creative AI, Mandinka, and N'Ko

TikTok Series Plan: Creative AI, Mandinka, and N'Ko

This series should not feel like a research paper cut into clips. It should feel like a creator showing people a completely different way to use AI.

The public premise is:

> I used AI to reconnect with my native language, Mandinka, and it turned into a research project about whether machines can hear and write the language in the script built for it.

That is the center. The audience does not meet the project through CER, CTC, TAR, TTT, AGP, or paper structure. They meet it through a surprising claim: AI is not only for making images, writing emails, or generating captions. It can become a microscope for culture, language, memory, and identity.

The technical work still matters. It gives the story seriousness. But on TikTok, the research is the receipt, not the opening line.

The Positioning

Most AI content says some version of:

> "Here is a tool that saves you time."

This series should say:

> "Here is a way I used AI to investigate my own language, my own script, and the infrastructure gap around it."

That difference is the hook. It is unusual. It feels creative without being gimmicky. It lets the viewer see AI as a research partner, a cultural mirror, a language-learning tool, and a machine that can be inspected when it fails.

The confidence should come from specificity:

> "I did not just ask AI about Mandinka. I used AI to test whether AI could actually represent Mandinka through N'Ko."

> "I used AI like a microscope, not a chatbot."

> "I wanted to know whether the machine could learn what I was learning."

> "This started with my native language. It became a question about who gets real infrastructure in AI."

What The Audience Should Feel

The viewer should feel three things in the first week:

First, this is personal. Mandinka is not an abstract dataset. It is the creator's native language, and N'Ko is a way into the deeper structure of that language.

Second, this is creative. The project uses AI in a way most people have not seen: scanning model internals, building a writing-system bridge, testing speech recognition, and turning language learning into an experiment.

Third, this matters. If AI cannot properly represent a language or script, then the future of voice interfaces, search, education, archives, and translation will silently leave people out.

The Series Promise

The promise should be simple enough for TikTok:

> I am going to show you how I used AI to learn, test, and build around my native language, Mandinka, through N'Ko.

A sharper version:

> I am using AI to study the language I come from, and the deeper I go, the more it shows how much modern AI is missing.

The audience is not being asked to understand the paper immediately. They are being asked to follow a journey that becomes more technical over time.

The Content Ladder

Do not start with the benchmark. Start with identity and curiosity.

The 20.57 result belongs in layer 6. It should not lead the series. It becomes powerful after the viewer understands why the experiment exists.

Core Pillars

Pillar 1: AI As Creative Research Partner

This is the broadest hook. It catches people who do not know Mandinka or N'Ko yet.

Use this pillar to show that AI can be used to ask unusual, personal, ambitious questions.

Example angles:

> "Most people use AI to generate content. I used it to investigate whether AI could understand my native language."

> "This is one of the strangest ways I have used AI: I made it help me study a script, then tested whether it could learn that script itself."

> "AI is more creative than people think when you stop treating it like a shortcut and start treating it like an instrument."

Pillar 2: Learning Mandinka Through N'Ko

This is the human anchor. It keeps the series from sounding like detached ML research.

Use this pillar to talk about the act of learning: characters, sound, tone, right-to-left reading, mistakes, patterns, recognition, and the feeling of seeing your own language in a script designed for it.

Example angles:

> "I grew up around Mandinka, but learning N'Ko made me look at the language differently."

> "The script forced me to pay attention to sound in a way Latin does not."

> "Every character started feeling like a small engineering decision."

Pillar 3: AI Blind Spots

This is the conflict. It turns the series from a learning diary into a larger argument.

Use this pillar to show that AI support is often shallow. A script can appear on screen while remaining weak inside the model.

Example angles:

> "The machine could show N'Ko, but when I checked inside the model, the signal looked weak."

> "Unicode is not understanding."

> "A language can be visible to your phone and still invisible to the model."

Pillar 4: Building The Machine

This is the builder arc. It shows competence and gives the audience a reason to trust the work.

Use this pillar to explain the pipeline visually: audio, Whisper features, decoder, N'Ko output, bridge, correction, row review.

Example angles:

> "Instead of asking AI to translate Mandinka after the fact, I wanted the model to write N'Ko directly."

> "The bugs in the bridge taught me where Latin hides sound."

> "The model was not only learning audio. It was learning the writing system."

Pillar 5: The Bigger Future

This is where the series becomes aspirational without becoming vague.

Use this pillar to connect the project to language infrastructure, African scripts, education, archives, voice tools, and creative AI.

Example angles:

> "The future of AI should not require every language to squeeze itself through English or French."

> "If your language is not represented inside the model, you are not fully included in the tool."

> "This is about making AI meet the language where it is, not forcing the language to meet AI where it is."

First-Week Posting Order

The first week should not explain everything. It should make people want the explanation.

Hook Bank

Use hooks like these. They are TikTok-native, not paper-native.

> "I used AI to study my native language, and it exposed something weird about modern AI."

> "This is probably the most personal way I have ever used AI."

> "Most people use AI to make content. I used it to test whether machines can understand where I come from."

> "I started learning N'Ko, and then I accidentally turned it into an AI research project."

> "This script was designed in 1949, and it might be better suited for speech AI than the systems we keep forcing onto it."

> "AI can display this script. That does not mean AI understands it."

> "I used AI like a microscope, and pointed it at my native language."

> "What if the problem is not that the language is hard, but that AI was never built for it?"

> "This is not a prompt hack. This is what happened when I used AI as a research instrument."

> "The deeper I got into N'Ko, the more I realized the AI problem was not just technical. It was infrastructural."

What To Delay

Delay these until the audience already understands the story:

Safe Way To Mention 20.57 Later

When the time comes, say:

> "After building the system, the strongest archived checkpoint reported 20.57

That frames the result as a receipt, not a hype claim.

Camera Rule

If a sentence sounds like it belongs in a paper, it is too early for TikTok.

If it sounds like:

> "Here is the weird thing I tried, here is what it showed me, and here is why it changed how I think about AI,"

then it belongs on camera.