Generate video with Google Veo 3 / 3.1 through Kie.ai's dedicated Veo endpoint. Veo's standout: native synchronized audio (dialogue, SFX, music) and strong prompt adherence.
Setup: needs
KIE_API_KEYin~/.claude/.env. Veo is premium — most expensive of the video skills (4K ~2× Fast). Mention this before generating.
Models (tiers)
| Key | model | Notes |
|---|---|---|
veo3 |
veo3 |
Veo 3.1 Quality — best fidelity |
veo3-fast (default) |
veo3_fast |
Faster, cheaper, great default |
veo3-lite |
veo3_lite |
Cheapest tier |
Flow
- Parse
$ARGUMENTS. Images change the mode automatically:- 0 images → text-to-video
- 1 image → image-to-video
- 2 images → first + last frames
- 3 images → reference-to-video
- Settings (ask, or defaults):
- Tier: default
veo3-fast. Suggestveo3for hero shots. - Aspect ratio:
16:9(default),9:16,Auto. - Resolution:
720p(default),1080p,4k. (Script auto-fetches the hi-res file for 1080p/4k.) - Duration:
4,6, or8seconds (default 8).
- Tier: default
- Craft the prompt — Veo rewards detail incl. audio cues (e.g. "she says: '…'", ambient sound, music). Show before generating.
- Generate:
python3 ~/.claude/skills/kie-veo-video/generate.py "<prompt>" \
--model veo3-fast --aspect-ratio 16:9 --resolution 720p --duration 8 \
[--image <url-or-path> ...] [--translate] \
--slug <slug>
- Present: output →
~/videos/veo/<date>-<slug>/. Note the audio is baked in. Offer Quality tier, 1080p/4k, or a vertical 9:16 cut.
Notes
- Veo audio is included by default — no separate step needed.
- 4K costs roughly double Fast; confirm with the user before large batches.
- Local input images are auto-uploaded.