Sora vs Runway vs Veo

If you searched “Sora vs Runway vs Veo,” you’re probably picking a tool to turn text or an image into a short AI video. Before any feature comparison matters, there’s a fact that changes the answer: OpenAI is shutting Sora down. The app is already gone, and the API follows in a few weeks. That’s not a knock on Sora’s output quality — it’s a straightforward reason not to build anything on it right now. This page covers what happened to Sora, then does the real comparison people searching this phrase actually need: Runway against Veo, on cost, limits, and what each is genuinely good at.

Sora is being discontinued — this isn’t a rumor

OpenAI announced on 24 March 2026 that it was discontinuing Sora, posting “We’re saying goodbye to the Sora app. To everyone who created with Sora, shared it, and built community around it: thank you” from the official Sora account [1]. The shutdown ran in two stages: the consumer app and web experience closed on 26 April 2026, and the developer API is scheduled to stop accepting requests on 24 September 2026 [2][3] — twenty days from when this page was last verified. ChatGPT’s own pricing page no longer lists Sora as a Plus or Pro benefit [4]. OpenAI’s stated reasoning was a redirection of compute and research focus toward world-simulation research for robotics and toward its coding and enterprise products, not a quality problem with the model itself [3]; reporting also ties the decision to weak engagement (downloads fell from a peak near 3.3 million to about 1.1 million by February 2026) and a $1 billion Disney content partnership that was dropped alongside it [2][5].

Two things are still true and worth knowing if you’ve already got API access. Sora 2 (sora-2) and Sora 2 Pro (sora-2-pro) remain callable through 24 September 2026, generating 16–20 second clips extendable to 120 seconds, at $0.10/second (Sora 2, 720p) and $0.30–$0.70/second (Sora 2 Pro, depending on resolution up to 1080p), half that on the batch API [6]. And the app’s signature feature, cameos — a one-time identity and voice verification that lets you (or, with consent, someone else) appear in generated clips, with the option to revoke access or delete any video featuring your likeness at any time — was a genuinely distinctive piece of product design [7], even if it’s now going away with everything else. Wrapper platforms like Higgsfield , which route prompts to Sora, Veo, and other backend models behind one interface, will lose the Sora option along with everyone else once the API closes. For a new project starting today, Sora isn’t a real choice — it’s history. The rest of this page is Runway vs. Veo.

What Runway and Veo actually are right now

Runway’s current flagship is Gen-4.5, announced 1 December 2025 and generally available via the API from 10 February 2026, which topped the Artificial Analysis text-to-video leaderboard at launch [8][9]. Runway isn’t only a model, though — the same subscription also gives you access to Veo 3.1, Seedance 2.5, Wan, and several other third-party video and image models inside one workspace, which makes it as much a multi-model creative suite as a single product [10]. Gen-3 Alpha Turbo and the original Gen-4 Aleph were sunset on 30 July 2026, with Runway pointing existing users toward Gen-4.5 or Gen-4 Turbo (still active) and toward Aleph 2.0, respectively [11] — so if you see pricing for “Gen-3 Alpha Turbo” anywhere, it’s stale, not “Gen-4 Turbo.”

Google’s current flagship is Veo 3.1, released 13 January 2026 and described on DeepMind’s own model page as “our leading video generation model,” with a cheaper Veo 3.1 Lite variant added for developers in April 2026 [12][13]. Google also unveiled Gemini Omni at I/O 2026 in May, an “any input, any output” model that generates video among other media and that this wiki’s own coverage described as becoming Google’s headline video-generation effort going forward [14] — worth flagging because Google now has two video-capable model lines under one roof, and it isn’t fully settled which one carries the “Veo” name a year from now. Treat Veo 3.1 as the current, documented product and Gemini Omni as the direction things are moving.

Sources: Runway pricing and API docs [9][15][16]; Google DeepMind and Gemini API pricing [12][13][17]. Google’s consumer prices above are from its EU pricing page in euros, which is the currency the official page itself displays [18]; US-dollar figures reported elsewhere put AI Pro near $20/month and AI Ultra between $100 and $250/month depending on tier — check Google’s own subscriptions page for your region before buying, and see this wiki’s AI subscription pricing 2026 for the fuller multi-vendor picture. All figures were verified against the sources above on 4 September 2026; both companies change pricing without much notice.

What each is genuinely best at

Runway is the editing-suite choice. Its real differentiators aren’t the base model, which trades the top benchmark spot with Veo release-over-release, but the tools built around it: Motion Brush lets you paint motion vectors onto specific regions of a frame to control exactly what moves and how; Director Mode gives parametric camera controls — dolly, orbit, crane, rack focus — without a physical rig; Act-Two drives a character’s performance from a reference video; and Aleph 2.0 does in-context video editing on existing footage — swap a product, change the lighting, remove an object — while trying to preserve everything you didn’t ask it to touch, on clips up to 30 seconds at $0.28/second via the API [19][20]. This is why Runway shows up in agency and post-production workflows more than the other two: it behaves like a professional tool with a model attached, not a model with a chat box attached.

Veo is the Google-ecosystem choice, and that shows up in two concrete places a comparison table won’t capture. First, native audio: Veo generates dialogue, sound effects, and ambient noise as part of the same generation, no separate audio pass required, which is a real workflow saving if you want a finished clip rather than a silent one to sound-design yourself [12]. Second, YouTube: Google has folded Veo into Dream Screen, the AI-background and AI-clip tool inside YouTube Shorts, so a creator already in Google’s ecosystem can generate Veo video without leaving the app they publish to [21]. If your output’s destination is YouTube specifically, that’s a real, product-level advantage neither Runway nor Sora ever had.

Real limitations — of both, honestly

Runway’s costs scale in a way that’s easy to underestimate. The credit system looks cheap at the entry tier, but professional formats add up fast: ProRes or PNG-sequence export adds 5 credits/second, 10-bit/HDR adds 20, and anything over roughly 4 megapixels (near-4K) adds 40 credits/second on top of the base rate [15] — a short HDR clip can burn through a Standard plan’s entire monthly allowance in seconds. And as the table above notes, Runway’s own documentation doesn’t show native audio generation inside Gen-4.5 itself; multiple independent write-ups disagree on whether audio has since been folded into the model or still requires a separate ElevenLabs or Seed Audio pass, and this page could not resolve that conflict against Runway’s own docs — check the current model card before assuming synced dialogue works out of the box.

Veo’s biggest weakness is one Google states about its own product: its model page acknowledges that “creating videos with natural and consistent spoken audio, particularly for shorter speech segments, remains an area of active development” [12] — the native-audio advantage above is real but not fully reliable yet, especially for lip-synced dialogue. The 8-second base clip length is also short compared to Runway’s up-to-10-second generations and Sora’s old 20-second (extendable) clips, and Google’s Flow-credit system, like Runway’s, meters video generation against a shared pool that also covers images and other AI features, so it’s easy to burn through a plan faster than the sticker price suggests.

Which one to pick

If you only take one thing from this page: don’t start a new project on Sora, whatever you’ve read elsewhere that assumes it’s still an option. Between the two that remain, reach for Veo when audio and a Google/YouTube workflow matter more than editing control, and reach for Runway when the opposite is true. Neither is a wrong answer for a generic “AI video” need — they’re built for different jobs, and the job is what should decide it, not which one is more famous.

What this page can’t fully resolve

Google’s positioning of Veo against its newer Gemini Omni model is genuinely unsettled as of this writing — DeepMind’s own site still calls Veo 3.1 its leading video model, while this wiki’s separate coverage of Google’s 2026 generative-media releases treats Gemini Omni as the emerging headline product [14]. Which name Google leads with a year from now isn’t something this page can predict. Whether Runway’s Gen-4.5 has gained native audio since its initial release is also unresolved — sources disagree, and Runway’s own API changelog does not settle it either way. Both companies also change consumer pricing and credit allotments without much notice, so treat every number above as a snapshot from 4 September 2026, not a standing fact — check the linked official pages before you subscribe.

Further reading

• AI subscription pricing 2026 : the fuller comparison of consumer AI subscriptions across vendors, including Gemini’s other tiers.
• Higgsfield : a creator-facing platform that wraps Veo and other frontier video models behind motion presets — relevant if you don’t want to pick a model directly.
• Google Gemini : this wiki’s tool page for the app Veo ships inside.
• Programmatic video : the code-driven alternative to generative video, for when you need exact, repeatable output rather than a generated clip.
• What is generative AI? : a plain-language primer if text-to-video is your first generative AI tool.
• Remotion vs FFmpeg : video processing approaches, for after you have footage rather than for generating it.

Sources

1. Sora (official account), announcement of app discontinuation, 24 March 2026: https://x.com/soraofficialapp/status/2036546752535470382
2. TechCrunch, “OpenAI’s Sora was the creepiest app on your phone — now it’s shutting down” (24 March 2026), including download and revenue figures: https://techcrunch.com/2026/03/24/openais-sora-was-the-creepiest-app-on-your-phone-now-its-shutting-down/
3. The Decoder, “OpenAI sets two-stage Sora shutdown with app closing April 2026 and API following in September,” on the 26 April 2026 and 24 September 2026 dates and OpenAI’s stated compute-reallocation reasoning: https://the-decoder.com/openai-sets-two-stage-sora-shutdown-with-app-closing-april-2026-and-api-following-in-september/
4. OpenAI, ChatGPT pricing page, checked 4 September 2026 for the absence of Sora as a listed Plus/Pro benefit: https://openai.com/chatgpt/pricing/
5. Variety, “OpenAI Will Shut Down Sora Video App; Disney Drops Plans for $1 Billion Deal” (25 March 2026): https://variety.com/2026/digital/news/openai-shutting-down-sora-video-disney-1236698277/
6. OpenAI Developer Platform, video generation guide and pricing (sora-2, sora-2-pro: duration, resolution, and per-second rates), fetched 4 September 2026: https://developers.openai.com/api/docs/guides/video-generation , https://developers.openai.com/api/docs/pricing
7. MacRumors, “OpenAI’s Sora App Adds Character Cameos and Video Stitching Tools” (30 October 2025), on cameo consent and revocation controls: https://www.macrumors.com/2025/10/30/openai-sora-app-character-cameos-video-stitching/
8. CNBC, “Runway rolls out Gen 4.5 AI video model that beats Google, OpenAI” (1 December 2025): https://www.cnbc.com/2025/12/01/runway-gen-4-5-video-model-google-open-ai.html
9. Runway, “Introducing Gen-4.5,” official research announcement: https://runway.com/research/introducing-runway-gen-4.5
10. Runway Developer Docs, “Available AI Models” (Gen-4.5 alongside Veo 3.1, Seedance, Wan, and other third-party models on the same platform), fetched 4 September 2026: https://docs.dev.runwayml.com/guides/models/
11. Runway Developer Docs, API changelog, on Gen-4 Turbo and Gen-3 Alpha Turbo retirement (30 July 2026): https://docs.dev.runwayml.com/api-details/api_changelog/
12. Google DeepMind, official Veo model page (Veo 3.1 description, native audio, resolution, access points, and the spoken-audio limitation quoted above), fetched 4 September 2026: https://deepmind.google/models/veo/
13. Google, “Veo 3.1 Ingredients to Video” and Veo 3.1 Lite model card: https://blog.google/innovation-and-ai/technology/ai/veo-3-1-ingredients-to-video/ , https://deepmind.google/models/model-cards/veo-3-1-lite/
14. This wiki, Google’s Generative Media in 2026 : Nano Banana 2, Veo 3.1 Lite, Lyria 3 Pro, and Gemini Omni’s introduction at I/O 2026.
15. Runway Developer Docs, “API Pricing & Costs” (credit rates for gen4.5, professional/HDR format surcharges), fetched 4 September 2026: https://docs.dev.runwayml.com/guides/pricing/
16. Runway, official pricing page (Free/Standard/Pro/Max/Enterprise plans and monthly credit allotments), fetched 4 September 2026: https://runway.com/pricing
17. Google AI for Developers, Gemini API pricing (Veo 3.1 Standard, Fast, and Lite per-second rates by resolution), fetched 4 September 2026: https://ai.google.dev/gemini-api/docs/pricing
18. Google, Gemini subscriptions page (Free, AI Plus, AI Pro, AI Ultra pricing and Flow-credit allotments), fetched 4 September 2026: https://gemini.google/subscriptions/
19. Runway Research, “More Control, Fidelity, and Expressibility” (Motion Brush and Director Mode camera controls): https://runway.com/research/more-control-fidelity-and-expressibility
20. Runway, “Introducing Aleph 2 and Edit Studio,” and the Aleph 2.0 product page (in-context editing, input limits, API pricing): https://runway.com/news/introducing-aleph-2-and-edit-studio , https://runway.com/product/aleph-2
21. Google, “Veo 2 is coming to Dream Screen on YouTube Shorts”: https://blog.google/feed/veo-2-dream-screen-youtube-shorts/