Character-recognition accuracy of rendered Buena Mono at small sizes · ← all reports
Generated against Buena Mono v1.228 · last updated 2026-07-22
The modern OCR-B question (Frutiger 1967): how reliably do machines read this typeface? Each test line is rendered at small pixel sizes and recognized with Tesseract (LSTM, --psm 7) and a vision-language model; scores are character-level similarity to ground truth.
O0o 1lI| 5S 8B rn m cl d 2Z 9gqif (x1 != y0) { return b || l1; }const O0 = 'I1l|'; let S5 = "B8";printf("%d", 0xDEADBEEF ^ 0b1010);wget https://example.com/~user/v1.221| Instance | Size | L1 | L2 | L3 | L4 | L5 | Mean (Tesseract) | Mean (VLM) |
|---|---|---|---|---|---|---|---|---|
| Thin | 9px | 73% | 79% | 82% | 88% | 95% | 83% | 100% |
| Thin | 12px | 75% | 91% | 84% | 88% | 99% | 87% | — |
| Thin | 16px | 82% | 86% | 88% | 88% | 100% | 89% | — |
| Thin | 24px | 89% | 80% | 88% | 91% | 100% | 90% | — |
| Regular | 9px | 70% | 83% | 82% | 84% | 94% | 82% | 100% |
| Regular | 12px | 72% | 89% | 91% | 88% | 97% | 87% | 100% |
| Regular | 16px | 89% | 86% | 91% | 88% | 97% | 90% | — |
| Regular | 24px | 89% | 84% | 94% | 90% | 97% | 91% | — |
| ExtraBold | 9px | 80% | 77% | 85% | 71% | 89% | 80% | 100% |
| ExtraBold | 12px | 79% | 74% | 79% | 85% | 95% | 82% | — |
| ExtraBold | 16px | 89% | 76% | 93% | 81% | 92% | 86% | — |
| ExtraBold | 24px | 90% | 83% | 94% | 79% | 95% | 88% | — |
| Italic | 9px | 74% | 77% | 79% | 90% | 92% | 82% | — |
| Italic | 12px | 75% | 84% | 85% | 94% | 92% | 86% | — |
| Italic | 16px | 82% | 81% | 91% | 91% | 94% | 88% | — |
| Italic | 24px | 83% | 75% | 91% | 81% | 93% | 85% | — |








VLM column appears when machine-readability-vlm.json transcriptions are present (model: Claude, transcribing the rendered images).