I forgot to save the draft so I let DeepSeek write the post for me. The general point remains
Subject: Feature Request: Custom User-Defined Languages & Style Transfers in Kagi Translate
Dear Kagi Team,
I’ve been thoroughly enjoying the new Kagi Translate app. The interface is slick, and the translation quality is fantastic. However, while playing with the "fun languages" (like Corporate Jargon and Pirate Speak), I had a thought: Why stop at pre-made options? Let us create our own languages.
I am formally requesting the ability to define and share custom "style transfer" languages on the Kagi Translate platform. Here is why I believe this feature would be a game-changer and a perfect fit for the Kagi philosophy.
- The Fun of Customization is Endless
The "Corporate Jargon" option isn't just a translation; it is a performance. The entertainment value comes not from the literal words, but from the contextual shift. If I can create a custom language—say, "Philosopher's Absurdism," "90s Sega Localization," or "Sassy Southern Grandma"—the potential for entertainment and creativity becomes infinite.
It’s entertaining because:
· It reveals the hidden biases and tones in our own writing.
· It acts as a "literary instrument" that lets us compose existing text into new genres.
· It allows for private inside jokes (e.g., translating a to-do list into the style of a Shakespearean tragedy).
- Pre-Made "Fun" Options Aren't Enough
While "Corporate Jargon" is a great proof-of-concept, limiting the user to developer-curated options is restrictive. One user's "Hacker Slang" is another user's "Middle Earth Elvish." By allowing users to define the style rules (prompt structures, synonyms, and tone vectors), Kagi moves from being a simple tool to being a creative sandbox.
We shouldn't just be passengers on this ride; we should be the ones building the rollercoasters.
- Why General Purpose LLMs Fail Here
You might argue that I could just prompt ChatGPT to do this. However, general-purpose LLMs are "loose" with style. They summarize and paraphrase rather than translate literally.
When I ask ChatGPT to translate a biology textbook into "Gamer Jargon," it usually summarizes the info and adds a few "GG" and "Nerf" mentions. It loses the literalness of the original text.
Kagi Translate can do better because:
· It is designed for translation, not generation. It can preserve syntax, sentence structure, and specific keywords.
· It can allow users to map specific words to specific replacements (e.g., "ecosystem" -> "meta", "photosynthesis" -> "farming") while keeping the rest of the grammar intact. This is a strict "style transfer," not an AI hallucination.
- The "Freemium" / Commercial Aspect
This is where it gets interesting. I recognize that computational power isn't free, especially for complex style transfers.
My proposal: Provide "Custom Language Translation" as a premium tier (perhaps with a daily free tier for tinkering). This allows users to:
· Recreation: Translate a recipe into "Surfer Dude" for a laugh.
· Serious Work: Translate a rough bulleted list into a fully formatted "Business E-mail" style for professional communication.
This isn't a novelty feature; it is a productivity hack. Being able to instantly switch the register of a text without rewriting it entirely is extremely valuable for professionals.
In short: Kagi Translate has the potential to be the first "Expressive Translation Engine." Please give us the keys to the kingdom and let us craft our own linguistic filters.
I’d be happy to help beta-test this feature if you decide to build it!
Thanks for listening,
CCantThinkofanUsername
The custom languages remind me of primitive text generators like Postmodernism Generator (https://www.elsewhere.org/journal/pomo/) and the later New-Age Bullshit Generator (https://sebpearce.com/bullshit/) which generate nonsense based on a certain jargon (ie postmodernism or New Age terminology). It's mostly entertaining because it gives a map of knowledge in unexpected ways, uses jargon meaninglessly and could even serve as apart of a story. (I remember having a Custom Language that generated greentexts and I even had a prompt, lol).
