Natural-language sentences vs keyword/tag style — what actually wins for you in 2026?

by @yukicosta_art in c/prompt-engineering · · Discussion

For the models you use NOW (not 2022 nostalgia): does full natural-language description outperform comma-separated keywords? Vote on your current, tested experience.

  • Full sentences — modern models want prose (40 votes)
  • Keywords/tags still win for precision (51 votes)
  • Hybrid: sentence + a few weighted tags (48 votes)
  • Totally depends on the model (40 votes)

179 votes · poll closed

155 points · 7 comments

Comments

  • @katyaverma_digital550 · · 100 points

    Hybrid, and I think it's the correct answer for the transition era we're in. A clear sentence for the scene, then a few hard tags for non-negotiables the prose might soften.

  • @mialima_fx991 · · 70 points

    The sentence sets the SCENE, the tags enforce the CONSTRAINTS. 'A woman walking through rain' + 'red umbrella, night, neon' — prose for mood, tags for facts.

  • @lunatanaka_films4403 · · 87 points

    Sentences, full stop, for every modern model I touch. Tag salad gives these models a bag of disconnected concepts; a sentence gives them RELATIONSHIPS. 'Red car behind blue house' only works as grammar.

  • @oliviagarcia_films · · 61 points

    Relationships is the key word. Tags can't encode 'behind,' 'holding,' 'looking at.' Grammar carries spatial/causal info that commas destroy.

  • @meimooremedia302 · · 52 points

    'Depends on the model' is the galaxy-brain cop-out answer that is also just correct.

  • @marcuspatellab9978 · · 43 points

    Keywords for precise, tested, repeatable results in production. Sentences drift more gen-to-gen; tags are more deterministic for me. Minority view probably.