Real outputs from the public examples of this model on Replicate, with the prompt and settings that produced them. They were not generated live on this page.
Input
Prompt
the woman is giving an interview for a podcast, wearing a pink top with the logo, it also neatly says "Veo 3.1", she is in a midcentury modern studio with pink lighting, she talks about using Veo 3.1 with reference images to put things into videos you're making, the logo is also in a framed picture against black behind her
the woman are having a conversation in a coffee shop, with the logo in the background. They talk about using Veo 3.1 with reference images to put things into videos
Google DeepMind, formed in 2023 from the merger of Google Brain and DeepMind under Demis Hassabis, runs the Veo video-generation programme. After the headline launch of Veo 3 (May 2025, Google I/O) with native audio, the team shipped Veo 3.1 in late 2025 as an incremental quality and capability upgrade. Veo 3.1 improves prompt adherence, motion physics, character consistency across extended sequences and audio fidelity (including richer ambient soundscapes and more controllable dialogue). The model is exposed via Vertex AI, the Gemini API, Google Labs (VideoFX, Whisk) and the Flow filmmaking surface aimed at professional creators.
Latent video diffusion / DiT with joint audio-video diffusion and reference-frame conditioning
Veo 3.1 retains the joint audio-video diffusion architecture introduced in Veo 3: video is encoded into a spatio-temporal latent space via a 3D causal VAE and denoised by a transformer-based diffusion model, while a coupled audio diffusion module generates synchronized music, ambient sound and dialogue. Improvements in 3.1 are reported across motion physics (water, fabric, crowds), audio quality and the ability to stay consistent across extended sequences via reference-frame and subject-reference conditioning. Native clips run up to 8 seconds at 1080p with extensions for longer sequences and a separate cascaded super-resolution stage for 4K. Text conditioning uses Gemini-family encoders; image and reference-frame conditioning add identity and layout control. Training expands on the Veo 3 corpus with additional curated multilingual audio-video data and refined recaptioning.
All Veo 3 features plus improved physics, audio fidelity and consistency
Up to 8-second 1080p clips natively, 4K via cascaded super-resolution
Reference-frame and subject-reference conditioning
Joint audio-video diffusion with music, ambient sound and dialogue lip-sync
Rich cinematographic prompt vocabulary
Multilingual prompts via Gemini text encoders
Available via Vertex AI, Gemini API, VideoFX, Whisk and Flow
SynthID audio + visual watermarking
Best for: high-end commercial creative, longer sequences, branded campaigns with sound.
Entraînement et licence
Expanded curated multilingual audio-video corpus including licensed footage, public web video (including YouTube under Google's terms) and synthetic data, with multi-granularity captions from Gemini vision-language models.
Licence: Proprietary commercial licence via Google Cloud / Vertex AI and Gemini API; commercial use under Google's generative-AI terms; mandatory SynthID watermarking.
curl https://railwail.com/api/v1/videos/generations \
-H "Authorization: Bearer $RAILWAIL_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "veo-3-1",
"prompt": "A slow drone shot over a misty pine forest at sunrise"
}'
# Response: {"job_id": "...", "status": "queued", ...}
# Poll until status is completed, failed or cancelled:
curl https://railwail.com/api/v1/jobs/JOB_ID \
-H "Authorization: Bearer $RAILWAIL_API_KEY"
import os
import time
import requests
API = "https://railwail.com/api/v1"
headers = {"Authorization": f"Bearer {os.environ['RAILWAIL_API_KEY']}"}
job = requests.post(
f"{API}/videos/generations",
headers=headers,
json={
"model": "veo-3-1",
"prompt": "A slow drone shot over a misty pine forest at sunrise",
},
).json()
while True:
status = requests.get(f"{API}/jobs/{job['job_id']}", headers=headers).json()
if status["status"] in ("completed", "failed", "cancelled"):
break
time.sleep(5)
print(status["status"], status.get("output_url"))
const API = "https://railwail.com/api/v1";
const headers = {
Authorization: `Bearer ${process.env.RAILWAIL_API_KEY}`,
"Content-Type": "application/json",
};
const job = await fetch(`${API}/videos/generations`, {
method: "POST",
headers,
body: JSON.stringify({
model: "veo-3-1",
prompt: "A slow drone shot over a misty pine forest at sunrise"
}),
}).then((r) => r.json());
let status;
do {
await new Promise((r) => setTimeout(r, 5000));
status = await fetch(`${API}/jobs/${job.job_id}`, { headers }).then((r) => r.json());
} while (!["completed", "failed", "cancelled"].includes(status.status));
console.log(status.status, status.output_url);
Proprietary commercial licence via Google Cloud / Vertex AI and Gemini API; commercial use under Google's generative-AI terms; mandatory SynthID watermarking.
Kuaishou's Kling v2.1, generating 5 and 10 second videos at 720p or 1080p from text or an image. Known for cinematic camera work and realistic physical motion, available on Replicate via the official KwaiVGI account.
