Developers found a cheaper way to feed Fable 5 large context by showing it pictures of text.
Rohan Paul Twitter · Rohan Paul (@rohanpaul_ai) · 2026-08-03
Developers released pxpipe, a tool that cuts Fable 5 token costs by rendering dense text as PNG images, exploiting the model's vision capabilities to pack roughly 92K characters into a single fixed-cost image block.
Appears in
Extraction
Topics: llm-inference-costcontext-compressionvision-language-modelstoken-optimization
Claims
- pxpipe renders text content into PNG pages and sends them as image blocks, replacing high-cost text tokens with a mostly fixed vision token cost.
- A 1928×1928 image costs approximately 4,761 vision tokens but can hold roughly 92,000 characters of dense text.
- Fable 5 uses OCR-like vision skills to read the pixel-rendered text, so semantic meaning generally survives the conversion.
- The technique is lossy compression: Fable 5 may misread exact IDs, hashes, names, or strings, making it risky for byte-exact facts.
Key quotes
a 1928×1928 image costs about 4,761 vision tokens. The same page can hold roughly 92K characters, so dense code becomes cheaper.
Fable 5 may understand the gist while misreading exact IDs, hashes, names, or strings. That makes it useful for bulky background context, but risky for byte-exact facts.
The catch is that this is compression through vision, not lossless text storage.