Sora API Pricing and 2026 Shutdown

Sora Web/App vs Sora API Timeline

OpenAI has announced a staged discontinuation of Sora services:

• Sora Web and App: Stopped as of April 26, 2026
• Videos API, sora-2 and sora-2-pro: Scheduled shutdown September 24, 2026
• Related model snapshots: Expected to stop alongside the API shutdown
• Recommended replacement: Not currently listed in OpenAI's deprecation table

Do not assume that the Sora API is already offline — it remains accessible until September 24, 2026, but the shutdown date is firm. Plan accordingly for any existing integrations.

Current Sora API Models

As of June 15, 2026, the active Sora video generation models are:

• sora-2 — standard video generation, 720p output
• sora-2-pro — higher quality and resolution options including 1024p and 1080p

Both models are covered by the September 24, 2026 shutdown announcement. Do not start new long-term Sora API integrations.

Standard Sora API Pricing

The following rates were reviewed on June 15, 2026 from the official OpenAI pricing page. Rates are per generated second of video. Prices may change before the September 24, 2026 shutdown; check the live pricing page before billing estimates.

Audio is included in Sora 2 video output capability. Do not add a separate audio fee unless the API documentation explicitly describes an additional audio charge for the specific workflow in use.

Batch API Pricing

OpenAI also offers a Batch API for video generation. Batch jobs are processed asynchronously and are suited for offline render queues rather than interactive use. Batch pricing is currently half the standard rate:

Batch pricing is also subject to the September 24, 2026 shutdown. Do not assume all accounts or job types are immediately eligible for Batch access; eligibility and availability are controlled by OpenAI.

Cost Formulas

Standard video generation

standard_video_cost =
  price_per_second
  × requested_generated_seconds

Batch video generation

batch_video_cost =
  batch_price_per_second
  × requested_generated_seconds

Project-level estimate

estimated_project_cost =
  sum(standard_video_cost or batch_video_cost for all jobs)
+ sum(edit_job_costs)
+ sum(extension_job_costs)
+ sum(retries_or_duplicates_that_created_billable_usage)

The final term depends on actual Usage and Costs records. Do not assume all failed or retried jobs are automatically billable or free.

Worked Cost Examples

Sora 2, 720p standard

• 8 seconds: 8 × $0.10 = $0.80
• 20 seconds: 20 × $0.10 = $2.00

Sora 2 Pro, 720p standard

• 8 seconds: 8 × $0.30 = $2.40
• 20 seconds: 20 × $0.30 = $6.00

Sora 2 Pro, 1024p standard

• 8 seconds: 8 × $0.50 = $4.00
• 20 seconds: 20 × $0.50 = $10.00

Sora 2 Pro, 1080p standard

• 8 seconds: 8 × $0.70 = $5.60
• 20 seconds: 20 × $0.70 = $14.00

Batch example — 100 videos at 8 seconds each, Sora 2 720p

100 × 8 × $0.05 = $40.00

These are cost estimates using the rates reviewed on June 15, 2026. Do not treat them as price guarantees; rates may change before the September 24, 2026 shutdown.

Duration and Resolution

Current OpenAI video generation guidelines:

• Maximum generated video length: up to 20 seconds per job
• sora-2 currently supports 720p output
• sora-2-pro supports 720p, 1024p and 1080p
• 1080p is not available on sora-2
• Higher resolution and longer duration tasks typically have higher queue latency

Acceptable seconds and size values are governed by the current endpoint schema. Verify against the live API documentation before assuming all integer durations within the 20-second cap are supported.

Sora 2 vs Sora 2 Pro

The primary differences relevant to cost estimation:

• sora-2 offers 720p at $0.10/second — lower cost for standard-quality video
• sora-2-pro adds 1024p and 1080p at higher per-second rates
• Both models are deprecated and scheduled for the same shutdown date
• Do not choose sora-2-pro solely for availability assurance — both are going away on September 24, 2026

Async Video Job Lifecycle

Sora video generation uses asynchronous jobs. A typical workflow:

1. Submit POST /v1/videos to create a generation job
2. Persist the returned video_id or job_id
3. Initial status may be queued or in_progress
4. Poll GET /v1/videos/{video_id} or use a webhook for status updates
5. Webhook events include video.completed and video.failed
6. Download output via the content endpoint when status is completed
7. If status is failed, inspect the error code and message before retrying

Do not create a duplicate job upon client-side timeout without first querying the original job's status. Polling or webhook consumption does not itself incur video generation charges.

Generate vs Edit vs Extension

The Sora API supports multiple operations including text-to-video generation, image reference, character reference, edit, extension and Batch jobs.

The current OpenAI pricing page does not list a separate flat fee for edit or extension operations. Do not assume an edit or extension is free, discounted or charged at a fixed fee. Record the resulting job, duration, model and organization cost record.

Each extension operation can add up to 20 seconds of video. A single video can be extended up to six times, reaching a maximum total length of 120 seconds. Each extension creates a new billable job — do not estimate a 120-second video by multiplying $0.10/second by 120; count each intermediate extension as a separate job in your project total.

Failed, Cancelled and Moderation-Blocked Jobs

OpenAI's video generation documentation does not state a universal rule that all failed, cancelled or moderation-blocked jobs are billed or unbilled.

Do not assume that every failed, cancelled or moderation-blocked video job was billed or free. Record the video ID, status, error code and message, model, size and requested seconds, then inspect the organization Usage and Costs records before retrying.

Distinguish the HTTP status and error code — a rate-limit response and an account-quota response require different handling. Check the error body and the organization cost record rather than assuming a uniform policy.

Usage and Cost Reconciliation

For cost reconciliation:

• Reconstruct cost from per-second rates and generated durations in your job records
• Cross-reference with the organization Usage and Costs data in the OpenAI dashboard
• Identify and exclude duplicate or retry jobs that created unnecessary charges
• Track Batch vs standard usage separately if both are used
• Note the pricing review date for each estimate

September 2026 Migration Checklist

Given the confirmed shutdown date of September 24, 2026, begin migration planning now:

• ✓Do not architect new long-term dependencies on Sora video output or OpenAI video IDs as primary keys
• ✓Export and back up all existing video results and generation metadata while the API is still available
• ✓Save prompts, input references, model, size, seconds and job IDs for every generation
• ✓Download and store output files in your own object storage — do not rely on OpenAI content URLs persisting after shutdown
• ✓Decouple your provider adapter layer from Sora-specific fields so migration can swap providers without rewriting business logic
• ✓Run small comparison tests against alternative video generation APIs before the shutdown date
• ✓Test alternative providers for audio capability, maximum duration, resolution support and async task handling differences
• ✓Set an internal deadline at least two weeks before September 24, 2026 to allow buffer for edge cases

No Official Replacement Listed

OpenAI's current deprecation table does not list a recommended replacement for the Videos API, sora-2 or sora-2-pro. Do not substitute another OpenAI model as an official replacement without a formal announcement.

For alternative video generation APIs, see the Video Generation API Pricing guide for a cross-provider overview. No third-party provider offers guaranteed feature parity with Sora's specific output characteristics.

What to Log

For every Sora API job, record:

• Video ID / job ID and timestamp (UTC)
• Model (sora-2 or sora-2-pro)
• Model snapshot version, if applicable
• Requested duration in seconds
• Requested resolution and aspect ratio
• Standard or Batch
• Operation type: generate, edit, extension or other
• Number of input references (image, character)
• Job status (completed, failed, queued, in_progress)
• Error code and message, if applicable
• Retry count for this job
• Webhook event received, if applicable
• Local estimated cost before submission
• Actual cost from organization Usage and Costs record
• Output file storage location
• Migration status for post-shutdown archival

Official Sources Reviewed

The information on this page is based on OpenAI's official documentation reviewed as of June 15, 2026. Pricing, model availability and the September 24, 2026 shutdown announcement may change; verify current status in the OpenAI Developer Portal before making integration decisions.

• OpenAI API Docs: Pricing
• OpenAI API Docs: Video Generation
• OpenAI API Docs: Deprecations
• OpenAI API Docs: Sora 2 model reference
• OpenAI API Docs: Sora 2 Pro model reference
• OpenAI API Docs: Changelog
• OpenAI Help: What to know about the Sora discontinuation

Related Guides

Frequently Asked Questions