Brand-voice content generator: five formats, five moods, two models
A web app that writes posts, stories, captions, idea lists and newsletters in one brand voice. Pick a mood and a format, switch between GPT-4o and DeepSeek R1.
Task
Social posts and newsletters get written in between other work, by different people. So the brand sounds different every time. A general-purpose chatbot doesn't fix it: you re-explain the tone, audience and format in every chat, and the text still comes out bland and needs editing.
Solution
I built a Streamlit web app with the brand's character baked in: who the audience is, which words fit, and a list of banned phrases. You pick one of five moods, from gentle support to a call for change, and one of five formats: post, stories, photo caption, idea list or newsletter email. The text appears in a few seconds. Copy it with one button or download it as a file. The model switches right in the interface. GPT-4o handles final copy, and the free DeepSeek R1 via OpenRouter is there for drafts and experiments.
Results
- Speed: a finished text in a minute instead of half an hour on a draft
- One brand voice: tone and vocabulary are set in the app itself, so they don't depend on who's writing
- 25 combinations: five moods across five content formats
- Lower costs: drafts run on a free model, the paid one is kept for final copy
Try the demo
Pick a mood, a format and a topic, and generate a text yourself. The demo app works in Russian; I can build the same in English or Turkish.