Sora vs Veo pitted OpenAI's Sora 2, released September 30, 2025, against Google's Veo 3 and then Veo 3.1. In 2026 the contest ended without a winner on quality. OpenAI discontinued the Sora app and website on April 26, 2026, and the Sora 2 API on September 24, 2026, so of the two, only Veo can still be used. Below is how they compared while both were live, what each cost, and what former Sora users can switch to now.
Sora vs Veo at a glance
Status and prices as of September 2026. The Sora column describes Sora 2 as it last worked.
| Feature | OpenAI Sora 2 | Google Veo 3.1 | Toybox AI Video Creator |
|---|---|---|---|
| Released | September 30, 2025 | October 15, 2025 (Veo 3 on May 20, 2025) | Powered by Google Veo 3.1 |
| Status | App and website closed April 26, 2026; API closed September 24, 2026 | Available in Google Flow, the Gemini API, Google AI Studio and Google Vids | Available in any browser on a Pro plan |
| One generation | Up to 20 seconds in the API | 4, 6 or 8 seconds, extendable to 148 seconds in the API | 4, 6 or 8 seconds, one clip per run |
| Resolution | 1080p exports on Sora 2 Pro in the API; 720p and 480p renders otherwise | 720p by default; 1080p and 4K in the API for 8-second clips | 720p |
| Sound | Speech, sound effects and background soundscapes | Native audio, always on | Clips come with sound; you describe it in the prompt |
| Real people | Your own likeness through "characters" in the app; the API blocked real people | Adults only in image-to-video through the API | Real people only with their explicit consent, under Toybox's Terms |
| Price today | Not available at any price | 50 free Flow credits a day; Google AI plans from $4.99 a month | Currently 350 credits per clip, on a Pro plan at $9.99 a month |
What happened to Sora
OpenAI shut Sora down in two steps, according to its help center: the app and website went first, on April 26, 2026, and the API followed on September 24, 2026. OpenAI's Sora 2 launch page now carries the same April 26 notice. OpenAI notified developers on March 24, 2026 that every Sora 2 model and snapshot would leave the API on that September date, and its video guide now states that OpenAI offers no direct replacement.
If you made videos in Sora, OpenAI's help article points you to the Sora sunset page, where an Export button starts the download and an email arrives once the file is ready. OpenAI recommends exporting as soon as possible, will email before any final export window begins, and says it will permanently delete Sora data after that window passes.
ChatGPT Images still makes and edits pictures on every ChatGPT tier, and the can ChatGPT make videos guide covers what ChatGPT can still do for a video project.
Veo 3 vs Sora 2 on quality
While both were live, Veo 3 vs Sora 2 was close on independent votes. Artificial Analysis still lists both on its with-audio text-to-video board, which turns blind votes between pairs of clips into Elo ratings. As of September 28, 2026:
| Model | Text-to-video Elo (with audio) |
|---|---|
| Veo 3.1 | 1088 |
| Sora 2 (December version) | 1083 |
| Sora 2 Pro | 1059 |
Each score has a 95% confidence interval of about 7 to 10 points, so Veo 3.1 and the December Sora 2 are effectively tied, with Sora 2 Pro behind. The same pattern held at the end of 2025: when Runway launched Gen-4.5 on December 1, 2025, it cited the same leaderboard with Veo 3 at 1,226 and Sora 2 Pro at 1,206.
Each maker also published its own claims. OpenAI said Sora 2 modeled physics more faithfully, giving the example of a missed basketball shot that rebounds off the backboard instead of teleporting into the hoop. Google's own raters, shown 1,003 MovieGenBench prompts, preferred Veo 3.1 overall, though Google says it couldn't test image to video against Sora 2 Pro, which didn't accept realistic human images at the time. Treat both as vendor claims.
Sora vs Veo 3.1 on length, resolution and sound
Sora 2 made longer single clips. OpenAI's API guide lists 16 and 20 second generations for both Sora 2 and Sora 2 Pro, with 1080p exports (1920x1080 or 1080x1920) on Sora 2 Pro and 720p or 480p renders otherwise. A Sora video could also take six extensions of up to 20 seconds each, for 120 seconds in total.
Veo 3.1 makes 4, 6 or 8 seconds per generation, at 720p by default and 1080p or 4K for 8-second clips through the Gemini API. It goes longer by extension: 7 seconds at a time, up to 20 times, for up to 148 seconds. In Google Flow you build longer scenes by extending 8-second clips.
Both generate sound with the picture. OpenAI described Sora 2 as a video and audio model that makes background soundscapes, speech and sound effects. Veo 3.1's audio can't be switched off in the Gemini API, and Google's Veo page calls short spoken lines an area it is still refining.
People, likeness and Sora's characters
OpenAI built the Sora app around "characters": after a short one-time video and audio recording, you could drop yourself into any scene. The Sora 2 API was stricter: it refused to generate any real person, public figures included, and turned away input images showing human faces.
If characters were why you used Sora, Google now offers something similar. The Gemini app lets you set up an avatar that carries your voice and likeness into new AI videos, and in Google Flow you add your avatar to a prompt by typing @me. Flow's avatars aren't available in the EEA, Switzerland or the UK. In Veo's API, image-to-video, first-and-last-frame and reference-image requests allow adults only, and people generation is more limited in the EU, the UK, Switzerland and the Middle East and North Africa.
Sora vs Veo pricing
Sora vs Veo pricing is now a one-sided question, because Sora can't be bought. OpenAI's discontinuation article covers exports, data deletion and refunds.
Veo 3.1 prices, as of September 2026:
- Google Flow, free. 50 credits a day on any account. A Veo 3.1 Lite clip costs 10 credits and a Veo 3.1 Quality clip costs 100. Unused daily credits don't roll over.
- Google AI Plus. $4.99 a month, with 200 extra Flow credits and video in the Gemini app, where Gemini Omni now makes the clips instead of Veo.
- Google AI Pro. $19.99 a month, with 1,000 extra Flow credits.
- Gemini API. Billed per second of video, with no free tier.
- Toybox Video Creator. Currently 350 credits per clip at any length, about $3.50 at top-up prices, on a Pro plan that currently costs $9.99 a month for 1,000 credits.
The Veo 3.1 pricing guide works out the cost per clip for every Google plan.
Sora vs Veo 3: which is better now
Veo 3.1 is the one to use now, because Sora no longer runs anywhere. On blind votes the two finished level, so moving costs little on quality. What former Sora users lose is the 20-second single clip and the social app with characters. What they gain is extension up to 148 seconds and a free daily allowance in Flow.
Older Sora vs Veo comparisons, including forum threads written before the shutdown, weigh two live products. The comparison that matters now is Veo against the models that have since passed both on the same leaderboard. Seedance 2.0 scores 1210 and Gemini Omni Flash 1233 on text to video, and Kling 3.0 at 1080p scores 1095. The Veo 3 vs Kling comparison covers the closest of those matchups, and the Sora alternatives guide lists the main replacements with their prices and limits.
Which one should you pick
Since Sora is gone, the choice is which Veo route, or which rival, fits your work:
- You liked making short clips with sound and don't want to pay yet. Use Veo 3.1 on Google Flow's daily credits.
- You used Sora to put yourself in scenes. Use the Gemini app's avatar, or Flow's @me references.
- You need longer single clips than 8 seconds, or dialogue in Spanish, Japanese or Korean. Kling 3.0 makes up to 15 seconds and voices five languages, so it's the closer match to Sora's longer clips.
- You built on the Sora API. Move to the Gemini API for Veo 3.1, and budget per second of video.
- You want Veo 3.1 clips without a Google plan, alongside your Toybox images and flyers. Use Toybox Video Creator.
Move a Sora idea to Veo 3.1 in Video Creator
Video Creator is powered by Google Veo 3.1, so a prompt you wrote for Sora usually needs trimming more than rewriting.
- Cut the Sora prompt down to one shot that fits 8 seconds: one subject, one action and one camera move. A 20-second Sora scene becomes two or three Video Creator clips.
- Paste the shot into "Describe your video" (up to 2,000 characters). Keep the sound lines from your Sora prompt: clips come with sound, and the description is the only place to ask for dialogue or effects. Check spoken words afterward, since they aren't guaranteed to match exactly.
- If you saved a frame from an exported Sora video as an image, tap "Start from a photo" and use it as the start frame. Toybox doesn't crop or resize frames, so use an image in the same shape as the clip.
- Pick Wide (16:9) or Vertical (9:16) under "Format", then 4, 6 or 8 sec under "Length".
- Tap "Create video" and give it a few minutes. If the clip fails, Toybox refunds the credits.
Video Creator makes 720p clips at a current 350 credits each and needs a Pro plan; see Toybox pricing. A clip you don't save is kept for 24 hours, so save any take you plan to reuse.