Sora 2 pricing: discontinued, was $20

Sora 2 pricing is historical. The product is discontinued, so there is no live plan to buy. Use this page to understand what projects used to cost, then move the same budget to Veo 3.1, Kling 3, or Seedance depending on the job.

Updated August 2026: Sora 2 is no longer for sale. OpenAI discontinued the Sora app and web experience on 2026-04-26, and the Sora 2 API shuts down on 2026-09-24 — a little over a month from now. Reported cause: the product was burning roughly $1M/day in compute against about $2.1M in lifetime revenue, and Disney exited the partnership. Sora survives only as a generation feature inside ChatGPT Plus/Pro, with no API, no batch processing, and no pricing tier to budget against. Everything below this line — the tiers, the per-second math, the optimization tactics — is now a historical record of what Sora 2 cost while it was buyable, not a current pricing guide. If you're deciding where to put that budget instead, skip to what replaces Sora 2's spend below, or read the full creator survival guide and, if you still have a pipeline calling the Sora API directly, the migration plan before September 24 breaks it.

Section 1: What Sora 2 pricing tiers used to be

Before the shutdown, Sora 2 was bundled inside ChatGPT subscriptions rather than sold as a standalone API product for most users — there was never a public per-call API tier for individual creators; generation ran through the ChatGPT web app, mobile apps, and the dedicated Sora app, none of which still exist. Here is how the consumer pricing stacked up while it was live.

Tier	Monthly Price	Sora 2 Access	Resolution Cap	Duration Cap	Concurrent Jobs
Free	$0	None (since Jan 10, 2026)	n/a	n/a	n/a
ChatGPT Plus	~$20	Limited daily allotment	1080p	~20s per clip	1
ChatGPT Pro	~$200	High monthly allotment	1080p / 4K (selective)	~60s per clip	2-3
ChatGPT Enterprise	Custom	Negotiated	1080p / 4K	~60s	Custom

Plus subscribers typically saw a soft cap that reset daily, often described as "roughly 50 priority generations per month" plus throttled access after that. Pro subscribers got the bigger pool, faster queue priority, fewer watermarks on download, and access to longer extensions and higher-resolution exports. Enterprise pricing was negotiated and included data-handling guarantees that most agencies and studios required before using Sora 2 in client work. None of this is available to buy anymore — the app and web experience shut down 2026-04-26, and the underlying API follows on 2026-09-24.

Section 2: How Sora 2 pricing used to play out

Headline subscription numbers hid what mattered: cost per finished second of video you actually shipped. Below are three project budgets, preserved as a record of how Plus and Pro math used to work out — useful now mainly as a reference point for what you're replacing.

Project A — Solo creator, 5 short-form clips per week

A creator on ChatGPT Plus at $20/month generating roughly 20 video attempts per week (with maybe 5 keepers) is producing around 100 attempts and 20 finished 15-second clips per month. That works out to 300 seconds of usable footage. Effective cost: $20 / 300s = ~$0.067 per finished second, or about $1.00 per 15-second clip — before you factor regenerations.

Project B — Marketing team, 10 product videos per month

A team subscribing to ChatGPT Pro at $200/month, generating 10 polished 30-second product videos with an average of 6 regenerations per finished clip, consumes about 60 generations per month. Effective cost: $200 / 300 finished seconds = ~$0.67 per finished second. The math gets better as you push closer to the monthly cap.

Project C — Agency, 40 client videos per month

An agency burning through the Pro plan needs roughly 240 generations to ship 40 clean 30-second clips. They typically hit the Pro cap mid-month and either queue a second Pro seat ($200 × 2 = $400/month) or move clients to Enterprise. Effective cost lands around $0.33-$0.55 per finished second once volume is high.

Section 3: Hidden costs of Sora 2 (for the record)

The sticker price was never the real price. Here is what teams underestimated, kept for anyone auditing an old budget or writing the post-mortem on a Sora-based pipeline.

Regenerations. Sora 2 was excellent but not deterministic. Plan on 4-8 regenerations per usable clip for tightly art-directed work. If your monthly allotment was 100 generations, that was really 12-25 finished clips.

Watermarks on Plus. Plus exports carried a visible Sora watermark. Removing it required Pro or above. If you needed clean exports, Plus was effectively unusable for client work.

Audio limitations. Sora 2 generated synchronized audio, but voice cloning and bespoke voiceover required a separate pipeline. Most teams paired Sora with ElevenLabs v3 or a voice cloning tool — that was a second subscription. That pairing pattern still holds for whatever replaces Sora 2 in your stack — Kling 3.0, VEO 3.1 and Seedance 2.5 all generate native audio, but bespoke branded voiceover is still its own step.

Extension costs. Pushing a clip past the base duration (the 60s ceiling on Pro) burns multiple generations from your quota. A single 60s clip extended twice can consume 3-4x the credit weight of a base 20s clip.

Plan limits during peak hours. Even Pro subscribers saw queueing during US business hours. Time-sensitive deliverables could stall.

Storage and re-export. Sora's project library had retention limits, and OpenAI's own shutdown notice made this literal: exported projects stop being retrievable after the deprecation windows close. Anyone who left footage sitting in the app rather than archiving it externally is the cautionary tale here.

Section 4: Cost per output — Sora 2 vs the wider market, as it stood before the shutdown

For historical context, here are the rough mid-2026 cost-per-second figures across the dominant video models while Sora 2 was still a line item. Numbers are approximate and the Sora 2 rows are no longer purchasable at any price.

Model	Access Method	Approx Cost per Finished Second
Sora 2 (Plus) — discontinued	$20/mo subscription	~$0.05-$0.10
Sora 2 (Pro) — discontinued	$200/mo subscription	~$0.30-$0.70
VEO 3.1 (Vertex AI)	Per-second metered	~$0.50-$0.75
Runway Gen-4	Per-credit	~$0.30-$0.50
Kling 3.0	Subscription + credits	~$0.15-$0.35
PixVerse V6	Credit packs	~$0.10-$0.20
Wan 2.7 (self-host)	GPU rental only	~$0.02-$0.08

Sora 2 Plus had the cheapest effective cost per second of any major hosted model — but only if you stayed inside the allotment, and only while it was for sale. That's moot now. The rows worth budgeting against today are VEO 3.1, Runway Gen-4/4.5, and Kling 3.0, alongside newer entrants the market has added since — HappyHorse-1.0, Gemini Omni Flash, and Seedance 2.5 among them — none of which have a published per-second Versely rate captured in this table; check current model pricing for live figures.

What replaces Sora 2 in your budget now

If you had a monthly line item for Sora 2 — a Plus seat, a Pro seat, or an API allocation — here's where that money goes instead, without inventing numbers this post can't verify:

• Cinematic realism with dialogue: VEO 3.1 is the closest match to what Sora 2's Pro tier was buying — native audio, strong prompt adherence, premium output.
• Volume short-form and character consistency: Kling 3.0 (audio-native since its February 2026 release) is the workhorse replacement for the bulk of what a Plus subscription used to cover.
• Joint audio+video at the frontier: HappyHorse-1.0 is the current #1 on Artificial Analysis's no-audio image-to-video and text-to-video leaderboards, generating audio and video in a single pass.
• Long single-take, high-resolution work: Seedance 2.5 does native 30-second generation at up to 4K with native audio — well past anything Sora 2's duration caps allowed.
• Conversational, iterative editing: Gemini Omni Flash is built for multi-turn video edits rather than one-shot generation.
• Self-hosted / open-weight: Wan 2.7 ships under Apache 2.0. LTX-2.5 is open-weight under Lightricks' own Community License — free for commercial use under $10M in trailing annual revenue, paid above that line — not Apache-licensed, so check the terms before assuming it's unrestricted.

Exact current per-second and per-credit rates for all of these live on Versely's model catalog and pricing page, which update as providers change terms — a static number in a blog post ages badly, which is the whole lesson of this page.

Section 5: How Sora 2 spend used to get optimized

These were the tactics Sora 2 users ran to keep a subscription-capped tool predictable. They're preserved here because most still apply — pointed at whichever model has replaced Sora 2 in your stack — even though the specific Sora mechanics (Plus/Pro seats, daily caps) no longer exist to optimize.

Storyboard before you generate. A two-minute sketch session saves regenerations regardless of which model you're prompting. Specify shot framing, motion, lighting, and a clear subject reference every time.

Use shorter clips and stitch. Extended clips typically cost more per second than base-length ones on most credit-metered models. Several shorter clips edited together often gives more control for a similar total spend.

Tier your models, not just your seats. The old move was one Plus seat for ideation and one Pro seat for finals. The current equivalent: draft on a cheap model (Wan 2.7, LTX-class) and reserve a premium render (VEO 3.1, Kling 3.0, Seedance 2.5) for the shot that ships.

Generate b-roll on cheaper models. Use an AI b-roll generator running on Wan 2.7 or PixVerse for filler shots and reserve premium models for the moments that need their motion fidelity.

Batch overnight where queue priority still matters. Off-peak generation windows still reduce wait time on several hosted models — check whether your current provider's queue behaves the same way Sora 2's did.

Section 6: Common Sora 2 pricing mistakes (and the one that matters now)

Treating it as unlimited. Plus was never unlimited. Once the daily cap hit, you waited — or upgraded.

Forgetting the audio pipeline. Sora 2's native audio was fine for ambient, not for branded VO — budget for voice cloning and music regardless of which model generates the base clip.

Ignoring watermarks until export. Discovering on deadline that your Plus exports were watermarked was the most common avoidable mistake in the Sora 2 era.

Buying Pro for one project. A single hero campaign often didn't justify a full month of Pro. That logic still applies to any subscription-tier model today.

The mistake that matters now: still routing new work to Sora 2. The API shuts down 2026-09-24. Any pipeline still pointed at it needs a migration plan, not a pricing optimization. For a full model comparison, see our 2026 video model roundup — VEO 3.1, Kling 3.0, and Wan 2.7 each win specific use cases, and none of them disappear on a vendor's timeline the way a single-model bet on Sora 2 just did.

FAQ

Is there a Sora 2 API I can pay per-call for?

There never was a public per-call Sora 2 API for individual creators, and it's moot now regardless — OpenAI shuts down the Sora 2 API entirely on 2026-09-24. If you have an existing production integration, see the migration plan.

Can I still generate with Sora 2 at all?

Only inside ChatGPT Plus/Pro as a chat feature, with tight rate limits, short clips, and no API or batch access. For any production workflow, treat Sora 2 as unavailable.

Does Sora 2 cost more for 4K vs 1080p?

This was true while Sora 2 was purchasable: 4K exports consumed more of the monthly quota per generation, and effective cost per finished 4K second ran roughly 2-3x the 1080p equivalent. It's historical information now.

How does Sora 2 pricing compare to VEO 3.1, now that Sora 2 is gone?

Sora 2 had a lower entry cost ($20 Plus tier) but hard usage caps. VEO 3.1 is metered per second on Vertex AI, which scales linearly with volume instead of capping. See our Sora 2 vs VEO 3.1 capability comparison for the full breakdown — read as a historical-vs-current comparison now, not a live choice.

What hidden fees should I watch for on whatever replaces Sora 2?

Storage retention, audio pipelines, extension/render costs, and the gap between a subscription's included quota and your real demand — the same categories that used to catch Sora 2 users out apply to Kling 3.0, VEO 3.1, or whatever model you route to next. Plan for at least a 20% overhead on top of the headline price.

---

Sora 2 was genuinely category-leading for short-form generative video while it existed, but OpenAI discontinued it — the app and web experience shut down 2026-04-26, and the API follows 2026-09-24. If you want a single dashboard that bundles VEO 3.1, Kling 3.0, Seedance 2.5, Wan 2.7, and the rest of the current 2026 model lineup — with transparent per-generation pricing instead of a vendor's subscription quota that can vanish on four months' notice — try the Versely AI video generator and the AI movie maker. One bill, every model still standing.