Loading page...Загружаем страницу...正在加载页面...
AtlasRepo

Every rhyme rule I gave Gemma, it found a loophole. Catrina News turns today's tech headlines into Día de los Muertos calaveras

Description and use

An automated pipeline that pulls tech news, filters headlines, generates Spanish calaveras with Gemma, validates rhyme and syllable rules via Python, adds English glosses, and synthesizes audio with ElevenLabs.

Problem

Automating the creation of satirical Spanish poems (calaveras) from daily tech news while ensuring strict rhyme, syllable, and content rules.

Outcome

The pipeline consistently produces 10/10 rhyme compliance, 9/10 1‑3 rhyme, and 34/40 syllable compliance, with 59 s per calavera, and eliminates disallowed content.

Workflow

  • Pull last 24h Spanish tech news via SerpApi Google News engine
  • Filter headlines: drop real people, tragedies, violence, espionage, named nationalities
  • Prompt Gemma to generate up to three calavera candidates per headline
  • Run rima.py to check rhyme (lines 2 & 4), syllable count (7–9 per line), and banned words/sounds
  • If a stanza fails, provide feedback to Gemma and retry up to three times
  • Select best stanza per headline
  • Add line‑by‑line English gloss
  • Use voz.py to call ElevenLabs REST API and stitch audio clips
  • Publish as a Posada‑style broadside on catrinanews.tech

Evidence

pulls the last 24 hours of tech news with SerpApi's Google News engine

Source

lets Gemma pick the ones worth satirizing (and drops anything about real people or tragedies)

Source

writes a calavera for each, with every rhyme and syllable checked by plain Python, not by the model

Source

Gemma 3 12B + meter feedback + best‑of‑3: 10/10 rhyme 2‑4, 9/10 rhyme 1‑3, 34/40 syllables, 59 s

Source

  • Record date: 2026-11-02T05:29:13.394Z