How to Make AI Videos with Google Veo 3.1 on Soup.video (Tutorial)
Google Veo 3.1 is Soup.video’s photoreal cinematic workhorse — strong instruction following, synchronized audio lean, and flexible generation modes. This tutorial walks through every Veo workflow available on the platform.
Open the studio: Generate with Veo 3.1 →
Veo 3.1 capabilities on Soup.video
| Generation type | When to use it |
|---|---|
| TEXT_2_VIDEO | Pure prompt → clip |
| FIRST_AND_LAST_FRAMES_2_VIDEO | Control start and end composition |
| REFERENCE_2_VIDEO | Guide look with up to 3 reference images |
Resolutions: 720p, 1080p, 4K
Quality modes: Lite, Fast, Quality (fixed ~8s clips)
Veo 3.1 pricing on Soup.video
Approximate credit costs (1 credit ≈ $0.01):
| Resolution | Lite | Fast | Quality |
|---|---|---|---|
| 720p | ~$0.25 | ~$0.50 | ~$1.60 |
| 1080p | ~$0.30 | ~$0.70 | ~$1.60 |
| 4K | ~$1.00 | ~$1.60 | ~$2.40 |
Pro tip: Iterate prompts on 720p Lite, then re-run winners on 1080p Fast or Quality.
Credits: Pricing.
Step-by-step Veo 3.1 text-to-video
1. Visit /video/veo-3-1
2. Set Generation Type to TEXT
3. Write a cinematic prompt
Veo rewards precise lighting and camera language:
Wide establishing shot of a glass cliffside house at dusk, warm interior lights glowing, slow drone push-in through coastal fog, photoreal, natural color grade, 16:9
4. Choose resolution + quality
Start 1080p + Lite for most marketing drafts.
5. Pick aspect ratio
- 9:16 short-form
- 16:9 YouTube / landscape ads
- Auto when references dictate framing
6. Generate and evaluate
Check motion coherence, face stability, and whether the camera move matches the prompt. Adjust one clause at a time.
Advanced: first/last frame and references
First + last frame is perfect for product transitions (packaging → unboxing hero).
Reference mode keeps wardrobe, product design, or location consistency across takes.
Upload clean, well-lit images. Avoid heavy watermarks and busy collages.
Prompt formula that works on Veo
[Shot size] + [subject action] + [camera move] + [time of day / lighting] + [lens] + [mood]
Modifiers that help:
- Photoreal, natural skin texture, physically accurate shadows
- Anamorphic bokeh, 24mm wide, 85mm portrait
- Overcast softbox, golden hour rim light, practical neon
Avoid dumping brand slogans into the prompt unless you truly need on-screen text (Veo is better at scenes than tiny typography).
Veo 3.1 API example
curl -X POST https://soup.video/api/v1/generations \
-H "Authorization: Bearer sk_soup_your_key_here" \
-H "Content-Type: application/json" \
-d '{
"model": "veo-3-1",
"inputs": {
"prompt": "A cinematic tracking shot of a futuristic sports car driving through a neon city.",
"resolution": "720p",
"quality": "lite",
"aspect_ratio": "16:9",
"generationType": "TEXT_2_VIDEO"
}
}'
More: Generate docs · Models API.
Related guides
Generate your next photoreal clip on Veo 3.1 — pay only for the compute you use.