From Diffusion to TTF: Making an AI Font and Tackling Normalization

Read Articleadded Dec 3, 2025
From Diffusion to TTF: Making an AI Font and Tackling Normalization

After poor results trying to edit glyph vectors with an LLM, the author switched to a diffusion model that generated clean letter images. They converted images to SVG with PIL + Potrace and built a TTF, discovering that consistent typographic metrics (baseline, x-height, ascenders/descenders) are the real challenge. With grids, references, and scripting, results improved, pointing toward an accessible path to custom AI-made fonts.

Key Points

  • LLMs struggle with precise geometric edits to glyph vectors; text-only manipulation is unreliable for visual accuracy.
  • Diffusion models excel at generating legible, stylistically coherent letterforms from prompts on the first try.
  • A practical pipeline is: generate raster letters → threshold to black/white → vectorize with Potrace → assemble into TTF/OTF.
  • The main bottleneck is typographic normalization—maintaining consistent baseline, x-height, ascenders/descenders, and overshoot across glyphs.
  • Providing traced reference letters and a typographic grid improves consistency, but careful metric management remains essential.

Sentiment

Mixed, with a strong corrective stance regarding the 'world's first' claim. While some technical appreciation exists for the result, the discussion largely focuses on establishing prior art and delves into related topics like typeface copyright and the general challenges of font design.

In Agreement

  • The impressive result achieved with Nano Banana is acknowledged.
  • The difficulty and importance of achieving typographic consistency in font design are implicitly supported by comments suggesting AI needs more fundamental design understanding (e.g., playing a type design game).
  • The high cost of custom font creation (e.g., $2,000 per character) is implicitly accepted as a real market condition for bespoke typography.

Opposed

  • The article's implicit or explicit claim of creating the 'world's first' AI-generated font (or the first using Nano Banana) is strongly refuted, with commenters citing earlier projects and extensive research in the field.
  • The article is criticized for not showing the AI-generated font in practical use, which is deemed an 'obvious thing' to include.