Back to Details

fal-prompting

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

fal.ai Prompting

fal.ai 提示词编写指南

Runtime: All endpoint calls run via the genmedia CLI. See the 
genmedia
skill for command syntax; run 
genmedia init
once if not yet installed.
Model families have meaningful prompting nuances. A prompt that works on GPT Image 2 (long, structured, exact text in quotes) will fail on Happy Horse (which wants ~20 plain-English words). This skill collects family-specific guides.
运行环境： 所有端点调用均通过genmedia CLI执行。命令语法请查看
genmedia
技能；若未安装，请先运行
genmedia init
。
不同模型家族的提示词存在显著差异。适用于GPT Image 2的提示词（长篇、结构化、带引号的精确文本）在Happy Horse上可能完全无效（后者仅需约20个通俗易懂的英文单词）。本技能汇总了各模型家族的专属提示词指南。

When to load which reference

何时加载对应参考文档

ReferenceLoad when
kling.mdWorking with Kling video models (O3, v3), multi-prompt, element controls, Standard vs Pro tier
gpt-image-2.mdWorking with 
openai/gpt-image-2
or 
/edit
, structured prompts, EXACT TEXT, multi-image compositing
happy-horse.mdWorking with 
alibaba/happy-horse/text-to-video
or 
image-to-video
, brevity-first, camera language
参考文档加载场景
kling.md处理Kling视频模型（O3、v3）、多提示词、元素控制、标准版与专业版差异时
gpt-image-2.md处理
openai/gpt-image-2
或
/edit
端点、结构化提示词、精确文本、多图像合成时
happy-horse.md处理
alibaba/happy-horse/text-to-video
或
image-to-video
端点、优先简洁表述、使用镜头语言时

Universal principles (all families)

通用原则（所有模型家族）

  1. Visual facts beat prestige adjectives. Replace "stunning, cinematic, masterpiece" with "overcast daylight, brushed aluminum, 50mm feel."
  2. Style tags need visual targets. "Minimalist brutalist" → "cream background, heavy black sans serif, asymmetrical type block, generous negative space."
  3. One controlled variable per iteration. Comparison only works when one axis changes at a time.
  4. Inspect schema before assuming a control exists. 
    genmedia schema <endpoint_id> --json
    . Negative prompt, seed, multi-prompt, and reference-image fields differ across families.
  5. Per-family rules override universal advice. Happy Horse rejects what GPT Image 2 rewards.
  1. 视觉事实优于空泛形容词。将“惊艳、电影质感、杰作”替换为“阴天自然光、拉丝铝材质、50mm镜头视角”。
  2. 风格标签需搭配视觉目标。将“极简粗野主义”改为“米白色背景、粗体黑色无衬线字体、不对称文字块、充足留白”。
  3. 每次迭代仅控制一个变量。只有当单一维度变化时，对比测试才有意义。
  4. 先检查 schema 再假设存在控制项。运行
    genmedia schema <endpoint_id> --json
    查看。负向提示词、随机种子、多提示词和参考图像字段在不同模型家族中存在差异。
  5. 模型家族专属规则优先于通用建议。Happy Horse不接受的内容，GPT Image 2可能反而青睐。

Catalog cross-reference

目录交叉引用

For "which model do I use" questions, see 
fal-models-catalog
:
  • Text-to-image endpoint selection → fal-models-catalog/references/text-to-image.md
  • Text-to-video endpoint selection → fal-models-catalog/references/text-to-video.md
  • Image-to-video endpoint selection → fal-models-catalog/references/image-to-video.md
若遇到“做X任务用哪个模型”的问题，请查看
fal-models-catalog
：
  • 文本转图像端点选择 → fal-models-catalog/references/text-to-image.md
  • 文本转视频端点选择 → fal-models-catalog/references/text-to-video.md
  • 图像转视频端点选择 → fal-models-catalog/references/image-to-video.md