VEO 3.1 Video Generation

VEO 3.1 is Google’s flagship video model, and 4ALL API provides three modes: text-to-video, image-to-video, and reference-to-video, with multiple speed/quality tiers. Requests use a unified asynchronous task protocol: submit task → poll for result. See Video Generation (Async Protocol).

Model Overview

Models are named veo_3_1_{mode}_{tier}_{aspect ratio/resolution}:

Mode	Description	Representative Models
t2v	Text-to-video	veo_3_1_t2v, veo_3_1_t2v_1080p, veo_3_1_t2v_landscape
i2v	Image-to-video (first-frame image driven)	veo_3_1_i2v_s, veo_3_1_i2v_s_1080p, veo_3_1_i2v_lite
r2v	Reference image-to-video (character/style reference)	veo_3_1_r2v_fast, veo_3_1_r2v_fast_portrait

Tiers: standard (no suffix) has the highest quality; _fast is faster; _lite is the most economical; _s is the standard tier for image-to-video. Aspect ratio: _portrait / _landscape; resolution optional _1080p.

Submission Example

curl https://api.4allapi.com/v1/video/generations \
  -H "Authorization: Bearer $AIPROXY_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "veo_3_1_t2v_fast",
    "prompt": "无人机视角掠过清晨的热带海岸线,电影感调色"
  }'

For image-to-video/reference-to-video, include an image in the request (image field, URL or base64). See the protocol page for the polling and result retrieval flow.

Selection Recommendations

• Drafts and batch jobs: _lite / _fast; final renders: standard tier + _1080p;
• For vertical short videos, use _portrait models directly; do not generate a landscape video and crop it afterward;
• Billed per request, failures are refunded. See Model Marketplace for pricing by tier.