If you're picking an AI video model in 2026, two names dominate the conversation: Google Veo 3.1 and Runway Aleph. Both produce clips that were unthinkable a year ago. Both are expensive relative to image generation. And both have clear strengths that make one or the other the right call depending on what you're shipping.

This post pulls together what we've learned running thousands of jobs through each via our API gateway, so you can choose without burning credits on exploration.

TL;DR

  • Veo 3.1 Fast — best price-to-fidelity ratio for short text-to-video clips. ~30-60s generation. Strong camera motion and realistic lighting.
  • Runway Aleph — better at dynamic action, style consistency across shots, and creative direction. Slightly longer generation and higher per-second cost.
  • Both accept image-to-video. Both cap at ~10 seconds per clip in the current generation.

Price comparison (per second, 720p)

Running through AI Generate with a 40% default markup:

  • Veo 3.1 Fast (text-to-video): $0.42 per second
  • Runway Aleph: $0.77 per second

For a 5-second clip that's $2.10 vs $3.85 respectively. Over 100 clips the gap is $175 — enough to matter. Volume tiers bring both below 20% markup once you're spending $200+/month.

When to pick Veo 3.1

  • You need realistic natural scenes: people, nature, architecture, product shots
  • You care about cost and want to iterate on many variations
  • The action is simple or static — a dog running, a camera pan, a zoom
  • You don't need a specific artistic style baked in

When to pick Runway Aleph

  • You need complex action — dancers, sports, multiple agents interacting
  • You're chaining clips and want style consistency across them
  • The clip is part of a branded creative where style matters more than literal realism
  • You already know Runway's prompt conventions from Gen-3 work

How to call both from one API

The whole point of a gateway is you don't have to keep two sets of credentials. Here's the exact curl for each — same Bearer token, different model name:

curl https://aimarcus.eu/aigenerate/api/v1/jobs/createTask \
  -H "Authorization: Bearer sk-aig-..." \
  -d '{
    "model": "google/veo-3-1-fast",
    "input": {
      "prompt": "Golden retriever running through autumn leaves, cinematic, 5s",
      "aspect_ratio": "16:9"
    }
  }'
curl https://aimarcus.eu/aigenerate/api/v1/jobs/createTask \
  -H "Authorization: Bearer sk-aig-..." \
  -d '{
    "model": "runway/aleph",
    "input": {
      "prompt": "A dancer in motion, studio lighting, cinematic",
      "duration": 5
    }
  }'

Both return a taskId. Poll /api/v1/jobs/recordInfo?taskId=... every 3 seconds until state="success", or set a webhook URL in your account settings to get a push notification.

Which one wins?

For 80% of production workloads where you're generating short B-roll, product clips or social content — Veo 3.1 Fast. The price difference compounds, and the realism is on par for most non-action scenes.

For creative projects where style is the product — music videos, branded trailers, game teasers — Runway Aleph. The extra 80% in cost buys you the kind of output that Veo still struggles with.

Easiest way to decide: run the same prompt through both via the playground. At $1 per test, you'll have your answer in ten minutes.