Sora 2 vs Veo 3.1: availability, features, API and price

You’re choosing a video workflow, but comparing Sora 2 and Veo 3.1 as if both were new purchase options misses the first decision. Sora’s shutdown changes the task from picking a favorite demo to preserving existing work and finding an available replacement.

Watch Veo 3.1 examples

Google DeepMind’s Veo 3.1 demonstration introduces audio and creative controls. These are vendor-selected examples, not our side-by-side test.

Source: Google DeepMind: Veo 3.1 - Designed to empower creatives

This guide separates that access question from the older feature and price comparison. You’ll see what to check about Google’s offered route, supported inputs and the shot you need to make. Start with a live interface and a representative brief, then judge whether its output and delivery controls replace the part of Sora your project depended on.

Sora 2 vs Veo 3.1 at a glance

The table reflects the current buying decision. It does not claim that one model wins every prompt. Earlier Sora examples remain useful as historical output references, but they do not establish current access or future support.

The decisive difference is availability

For a new project, choose an available Veo 3.1 route rather than starting a Sora 2 dependency. OpenAI closed the Sora app and web experience on April 26, 2026 and published September 24, 2026 as the Sora API discontinuation date. Google’s Gemini API Veo 3.1 Preview models also shut down on October 22, 2026. Veo 3.1 Lite remains an option in Flow. Export existing work and compare the exact controls and price on the route you will use. See OpenAI's shutdown notice and Google's Veo documentation.

Sora 2 is a discontinued product surface. Existing users should request an archive at OpenAI's Sora export page and verify that the downloaded files include the outputs they need. Keep prompts, source images, settings and edit notes with the media so a replacement test starts from the same brief.

Veo 3.1 Lite remains available in Flow. Other providers may expose different Veo variants, inputs and resolutions. Check the active model and exact route before comparing output or pricing.

Google’s current Veo guide describes Veo-specific controls. Check which controls your host supports. For a direct Gemini API integration, Google’s the deprecation schedule lists the retired Preview IDs, and the current video guide directs new work to Gemini Omni 1.2 Flash or available Cloud GA models.

What can Veo 3.1 generate?

Google documents text-to-video and image-to-video generation with native audio. Veo 3.1 supports 16:9 and 9:16 output, four-, six- or eight-second generations, and 720p, 1080p or 4K output. Higher resolutions require eight seconds; 4K is unavailable on Lite. Video extension is limited to 720p and works with previously generated Veo video.

• Starting image: animate a supplied image as the first frame.
• First and last frames: constrain where an eight-second shot begins and ends.
• Reference images: guide a person, character or product with up to three asset images.
• Extension: add seven seconds to a compatible Veo-generated video, subject to Google's current limits.
• Audio: dialogue, effects and ambient sound are generated with the video and should be specified in the prompt.

These are model-level capabilities documented by Google. A browser app or third-party API can impose different limits. Confirm duration, resolution, references, safety handling, storage and download behavior on the interface you will actually use.

How much does Veo 3.1 cost?

The former direct Gemini API Veo 3.1 Preview rates are historical. For a new direct Gemini build, Google lists Gemini Omni 1.2 Flash at about $0.07 per second of 720p video output, before input charges. Check the current Google's current pricing and the quote for the specific Veo provider you plan to use.

At the listed direct-API rate, one eight-second Standard generation costs $3.20 at 720p or 1080p and $4.80 at 4K. One eight-second Fast generation costs $0.80 at 720p, $0.96 at 1080p or $2.40 at 4K. These calculations exclude retries, editing, storage and any third-party markup.

Budget with cost per accepted clip: total generation and correction spend divided by the number of outputs that satisfy the brief. A cheaper generation can cost more when it requires many retries or manual fixes.

How to run a fair Sora-to-Veo migration test

• Choose one recurring shot, not a showcase prompt.
• Preserve the original brief, source assets, aspect ratio, duration and audio requirements.
• Adapt prompt syntax only where the new provider documents a different control.
• Record every attempt, failure, charge, correction and accepted output.
• Review subject fidelity, motion, camera behavior, text, audio, continuity and policy fit separately.
• Scale only after the representative shot reaches an acceptable success rate.

For a product or character, test reference consistency. For a transition, test first-and-last-frame control. For dialogue, write the spoken line and sound events explicitly. This produces a decision tied to the workflow instead of a universal quality claim from one prompt set.

Which route should you choose?

Choose Veo 3.1 when its current provider route supports the inputs, resolution and delivery terms your project needs. Choose another available model when it produces a higher accepted-output rate for the same brief. Do not choose Sora 2 as a new durable API dependency after its announced discontinuation date.

You can compare a browser workflow through Magic Hour's text-to-video or image-to-video tools. Confirm the selected model and displayed settings before treating the result as a Veo test. For a broader replacement shortlist, use the Sora alternatives guide.