Google released Gemini 4 Hydrogen, a fast and cheap model. Useful for anyone who uses an LLM to expand prompts and write shot lists

by @diegoalmeidaworks312 in c/tutorials ·

Introducing Gemini 4 Hydrogen
Today, we’re introducing Gemini 4 Hydrogen, our daily driver for agentic workflows
blog.google

Not a video model, but relevant for how a lot of us work. Hydrogen is Google's new fast tier: $1.50 per million input tokens and $7.50 per million output, and Google says it finishes tasks about 2.5x faster than 3.8 Flash.

In my classes, most students use an LLM to turn a scene idea into a shot list and then into per-shot prompts. That job needs speed and consistency more than genius, so a fast cheap tier is a good fit.

If you've tried it for prompt expansion, how does it compare to what you were using?

#gemini #prompt-writing #llm #shot-lists

31 points · 5 comments

Comments

  • @dmitrischneider_motion784 · · 3 points

    At that price you can run three variations of every prompt and pick the best. Volume beats cleverness for prompt expansion.

  • @tess_story · · 3 points

    For shot lists I care about one thing: does it keep track of continuity (who's holding what, which side of the frame) across 20 shots. Fast models used to forget by shot 8.

  • @katyaitomedia512 · · 1 point

    Is it okay to use a cheaper model while learning, or will it teach me bad prompting habits?

  • @diegoalmeidaworks312 · · -1 points

    Cheaper is fine. The habit that matters is reading the output and editing it yourself, not which model wrote it.

  • @harpermoore_motion · · 0 points

    I've been using one long system prompt with my visual style rules, and any decent model follows it. The model matters less than the rules you give it.