Sora 2 vs Veo 3: Pricing and Migration After Sora's Shutdown

Google Veo 3.1 is the pick, because OpenAI shut the Sora API on September 24, 2026 and Veo 3.1 is the one of the two you can still buy.

At Layer3Labs, we map automation workflows for growing teams, and a model shutdown is the point where a team needs exact prices to plan the move.

Before its shutdown, third-party trackers reported Sora 2 Standard at $0.10 per second for 720p resolution, the same rate as Google Veo 3.1 Fast. Meanwhile, Sora 2 Pro's last reported 1080p rate of $0.70 per second sits above Veo 3.1 Standard's $0.40. Video-model prices change without notice, so check Google's Gemini API pricing before budgeting.

Sora 2 (OpenAI, discontinued) vs. Veo 3.1 (Google): Side-by-Side

Are you one of these vendors? Update your listing

Verdict on Sora 2 vs Veo 3: Google Veo by Default

In a direct Sora 2 vs Veo 3 comparison, Google Veo 3.1 wins by default because OpenAI shut the Sora API on September 24, 2026, after closing the Sora app and web experience on April 26. Developers evaluating both systems can only deploy Veo 3.1 today.

Google Veo is Google's video model, not the Veo sports-camera company. For teams that previously budgeted for Sora 2 Standard at $0.10 per second, Google Veo 3.1 Fast matches that unit rate directly at $0.10 per second for 720p video with native audio included. Veo 3.1 Lite reduces testing costs further to $0.05 per second at 720p.

If you require programmatic video generation today, Veo 3.1 offers working endpoints while Sora cannot generate anything.

Timeline of the OpenAI Sora Discontinuation

OpenAI announced the discontinuation of Sora on March 24, 2026, establishing a phased retirement schedule documented by outlets including NewsBytes. The initial phase took effect on April 26, 2026, when OpenAI closed the Sora app and web experience.

The final phase concluded on September 24, 2026, when OpenAI permanently closed the Sora API. Today, neither developers nor consumers can generate new media through any Sora interface. OpenAI directed former users to Codex for repurposing remaining ChatGPT and Sora credits, alongside standard subscription refund pathways.

Users seeking to preserve past generations can consult the OpenAI discontinuation help center for instructions on exporting media at sora.chatgpt.com/sunset. The documentation notes that data is subject to permanent deletion once export windows close. For a comprehensive walkthrough of the retirement schedule, read our guide on whether Sora is shutting down.

• March 24, 2026: OpenAI announced the Sora discontinuation.
• April 26, 2026: The Sora app and web experience closed.
• September 24, 2026: Sora API endpoints closed permanently for all developer accounts.
• Credits and refunds: purchased ChatGPT or Sora credits can still be used for Codex, and OpenAI's help page links ChatGPT subscription refund instructions.

Pricing Comparison and Usage Mapping

Because OpenAI removed its official pricing documentation, past Sora rates are referenced from third-party tracking data. According to records tracked on CostGoat, Sora 2 Standard was priced at $0.10 per second for 720p resolution. Sora 2 Pro commanded higher rates: $0.30 per second for 720p, $0.50 per second for 1024p (1792×1024), and $0.70 per second for 1080p high definition (HD).

Google structures Veo 3.1 pricing across three tiers in the Gemini API: Standard, Fast, and Lite. Veo 3.1 Fast at 720p costs $0.10 per second, creating an exact price parity with Sora 2 Standard. For 1080p output, Veo 3.1 Standard charges $0.40 per second, representing a lower unit cost than Sora 2 Pro at $0.70 per second. Veo 3.1 Fast delivers 1080p output at $0.12 per second, reducing unit costs even further.

An 8-second clip on Veo 3.1 Fast at 720p costs $0.80 (8 seconds × $0.10). That identical duration on Sora 2 Standard also totaled $0.80 (8 seconds × $0.10). At 1080p, an 8-second clip costs $3.20 on Veo 3.1 Standard (8 seconds × $0.40) or $0.96 on Veo 3.1 Fast (8 seconds × $0.12), whereas the same clip on Sora 2 Pro reached $5.60 (8 seconds × $0.70).

Veo 3.1 also offers a Lite tier at $0.05 per second for 720p and $0.08 per second for 1080p, yielding an 8-second generation cost of $0.40 and $0.64 respectively. For 4K production, Veo 3.1 Fast bills at $0.30 per second ($2.40 for 8 seconds) and Standard bills at $0.60 per second ($4.80 for 8 seconds). How extension seconds are billed is not covered in the pricing lines verified here, so confirm on Google's pricing page before planning clips longer than 8 seconds.

• Veo 3.1 Fast 720p ($0.10/s) maps directly to Sora 2 Standard 720p ($0.10/s).
• Veo 3.1 Standard 1080p ($0.40/s) operates below Sora 2 Pro 1080p ($0.70/s).
• Veo 3.1 Fast 1080p ($0.12/s) reduces 1080p generation costs to $0.96 per 8-second clip.
• Veo 3.1 Lite provides entry-level testing at $0.05/s for 720p ($0.40 per 8-second clip).
• Review our detailed analysis of Veo 3 pricing tiers for full budgetary breakdowns.

→ Compare per-token pricing for every AI model

Technical Specifications and Migration Constraints

Migrating an automated generation pipeline from Sora to Veo 3.1 requires adjustments to generation parameters and output handling. Google Veo 3.1 supports clip lengths of 4, 6, or 8 seconds per generation call. Requests generating 1080p or 4K resolution, or jobs using reference images, require the duration to be set strictly to 8 seconds.

Aspect ratios on Veo 3.1 are restricted to 16:9 widescreen or 9:16 vertical video at 24 frames per second. Workflows requiring square, 4:3, or cinematic 21:9 ratios must handle letterboxing or cropping externally. Audio generation is native and permanently enabled across all Veo 3.1 models, so every clip arrives with sound, and a pipeline that expects silent video has to strip or replace the audio.

Storage policies present another operational requirement. Google retains generated Veo video files on its servers for only 2 days before permanent deletion. Applications must immediately download and persist output files to internal object storage. Furthermore, every Veo 3.1 output carries a SynthID watermark.

Geographic and language constraints apply as well. In the European Union (EU), United Kingdom (UK), Switzerland, and Middle East and North Africa (MENA) regions, people generation is limited to adults (the allow_adult setting). English is fully supported, and Google notes that other languages have not been evaluated. Consult the Google Veo documentation to verify parameter support.

• Fixed clip lengths: 4, 6, or 8 seconds; 1080p, 4K, and reference-guided calls require exactly 8 seconds.
• Aspect ratios: 16:9 default and 9:16 vertical only; no native square or 21:9 modes.
• Data retention: Generated assets expire from Google servers after 48 hours.
• Audio handling: Native audio synthesis is active by default on every generation call.
• Safety and compliance: SynthID watermarking is mandatory; adult-only people generation applies in the EU, UK, Switzerland and MENA.

Alternative Models Beyond Sora and Veo

Engineers evaluating Veo 3.1 should note that Google's own documentation recommends Gemini Omni Flash as the default model for automated video generation. Model ID gemini-omni-1.1-flash is generally available on the paid tier of the Gemini API. It processes multimodal inputs including text, image, audio, and video, billing at 5,792 tokens per second of 720p video, which equals approximately $0.10 per second.

Google advises developers to reserve Veo 3.1 for specific advanced capabilities such as scene extension, last-frame control, or integration with legacy pipelines. For teams requiring wider aspect ratios or longer single generations, ByteDance offers the Seedance model family through BytePlus ModelArk. Seedance 2.5 generates clips from 4 to 30 seconds in a single call across aspect ratios like 21:9, 16:9, 4:3, 1:1, 3:4, and 9:16.

Pricing on Seedance differs from Veo. Official examples for a 5-second 16:9 clip show Seedance 2.5 costs $1.156 at 720p ($0.231 per second) and $2.843 at 1080p ($0.569 per second). This makes Seedance 2.5 more expensive per second than Veo 3.1 Fast at 720p ($0.10 per second) and Veo 3.1 Standard at 1080p ($0.40 per second). For detailed comparisons, see our guide on Seedance vs Veo.

Developers can also access models through multi-provider studios such as Runway, which has a free plan with 125 one-time credits and paid plans from $15 per month ($12 billed annually) for 625 monthly credits. Teams seeking broad ecosystem evaluations can review our roundup of Sora alternatives to compare additional commercial tooling.

• Gemini Omni Flash: Google's default recommendation at approximately $0.10/s for 720p video.
• ByteDance Seedance 2.5: Delivers up to 30-second generations on BytePlus ModelArk, costing $0.231/s at 720p.
• ByteDance Seedance 2.0: Offers 4K generation up to 15 seconds at $0.78 per second list price.
• Runway: free plan with 125 one-time credits, and paid plans from $15/mo for 625 credits with Gen-4.5 and Aleph 2.0.
• Explore our breakdown of Sora pricing history for historical context on early video APIs.

The Verdict

For any developer or business choosing between Sora 2 and Veo 3 in late 2026, Google Veo 3.1 is the sole operational choice. With OpenAI closing the Sora API on September 24, 2026, building on Sora is no longer possible. This would change only if OpenAI reopened Sora to buyers. Veo 3.1 does not suit a team that needs a free API tier or single takes longer than 8 seconds.

Migrating to Veo 3.1 Fast preserves unit economics for teams previously targeting Sora 2 Standard, matching its $0.10 per second 720p rate, with native audio always on. For 1080p, Veo 3.1 Standard at $0.40 per second and Fast at $0.12 both cost less than Sora 2 Pro's last reported $0.70.

Do not default to Veo 3.1 without testing alternatives. For standard production workloads, test Google's recommended Gemini Omni Flash first at roughly $0.10 per second, and evaluate ByteDance Seedance 2.5 via BytePlus ModelArk if your pipeline requires single generations longer than 8 seconds or flexible aspect ratios.