Hermes Video Generation Skill

Agent skill for generating short AI videos through FAL.

It supports:

• text-to-video;
• image-to-video;
• video/reference workflows;
• Russian/Cyrillic text overlays via ffmpeg after generation;
• presets for cheap, default, quality, seedance, kling, and wan.

Files

• SKILL.md — Hermes skill instructions.
• generate.py — CLI wrapper for FAL queue API.
• .env.example — environment template. Copy to .env and add your FAL key.
• requirements.txt — Python dependencies.

Generated media, downloaded references, .env, and cache files are intentionally ignored.

Setup

python3 -m pip install -r requirements.txt
cp .env.example .env
# edit .env and set FAL_KEY or FAL_API_KEY

ffmpeg is required for --overlay-text.

Usage

Dry-run without paid generation:

python3 generate.py "A bird flying across a blue sky" --preset default --duration 5 --dry-run

Text-to-video:

python3 generate.py \
  "A bird flying gracefully across a clear blue sky, realistic motion" \
  --preset default --duration 5 --aspect-ratio 16:9 --resolution 720p

Russian text overlay:

python3 generate.py \
  "A solemn Victory Day night sky, fireworks, no text" \
  --preset default --duration 5 \
  --overlay-text "С Днём Победы!" --text-position bottom --contrast-box

Image-to-video:

python3 generate.py \
  "Slow cinematic camera push-in, subtle natural motion" \
  --image ./reference.png --preset default --duration 5

Notes

• Prompts should usually be written in English for better video quality.
• Do not ask the video model to render Cyrillic directly; use --overlay-text instead.
• Video generation is paid. Start with --dry-run and short durations.
• OpenAI/Sora is not used by default; this skill is designed around FAL.