Correction, 2026-10-01: an earlier version of this page called the Sora shutdown "today" (it was published on 2026-09-06 and the shutdown was 2026-09-24), listed Wan 3.0 at $0.035, $0.07 and $0.14 a second (Alibaba's own page lists $0.05, $0.10 and $0.20), listed Veo 3.1 Fast at $0.15 a second (Google's price is $0.10 at 720p and $0.12 at 1080p; $0.15 is what Runway's API charges for it), said Sora's endpoints had disappeared from fal's catalog (fal lists them as deprecated), and said agents using our MCP server generate "on the canvas" (connector results come back as file links). Every vendor price below was re-read from the vendor's own pricing page on 2026-10-01.
The Sora 2 API was shut down on 2026-09-24, and OpenAI lists no replacement. So the practical question for anyone shipping video generation is not which model is best but which supplier you want to be exposed to next time this happens. The three routes are direct provider APIs (one relationship per model), aggregators like fal, Replicate and Runway (one integration, many models, prices that can differ from the provider's), and platform APIs that bundle a workflow. Disclosure up front: 8frame does not sell a raw generation API. We expose an MCP server, which is a different shape, and this guide says where each route wins including where we are the wrong answer.
TL;DR
- Sora 2,
sora-2-proand the Videos API stopped working on 2026-09-24. OpenAI's archived guide: they "were shut down on September 24, 2026 and are no longer available" - Price per call: direct provider APIs give you the provider's list price, at the cost of a separate integration, key, and billing relationship per model
- Fewest integrations: aggregators (fal, Replicate, and Runway, whose API now carries wan3, seedance2_5, grok_imagine_1_5, hailuo3, veo3.1, gemini_omni_flash and more)
- Per-second list prices (checked 2026-10-01): Veo 3.1 $0.40 Standard and $0.05 Lite (720p), Gemini Omni Flash about $0.10 at 720p, Wan 3.0 from $0.05, Grok Imagine Video $0.05, Runway Gen-4.5 $0.12
- If you want a workflow rather than an endpoint: an MCP server or a canvas, which is what we sell and what a raw-API team should not buy
The three routes
Direct provider APIs
You integrate Google, ByteDance, Alibaba, xAI or Runway individually. You pay the provider's own price, and you get features the moment the provider ships them.
The cost is operational: a key, a billing relationship, a rate-limit regime and a payload schema per provider, and a migration project every time one deprecates something. Which is exactly the Sora situation.
Aggregators
One integration, many models, and switching a model is a string change rather than a project. fal and Replicate are the established pair; Runway has become a third, since its developer pricing now lists wan3, wan3_prime, seedance2_5, grok_imagine_1_5, hailuo3, veo3.1, veo3.1_fast, gemini_omni_flash, seedream5_pro and more alongside its own models.
Do not assume a markup, and do not assume there is none: compare model by model. On Runway's API (credits at $0.01 each), Wan 3.0 at 1080p is 20 credits a second and Veo 3.1 with audio 40, the same $0.20 and $0.40 that Alibaba and Google list. Veo 3.1 Fast with audio is 15 credits, $0.15, against Google's own $0.10 at 720p and $0.12 at 1080p.
You also inherit their availability and their catalog decisions. fal's model catalog lists all six of its Sora 2 endpoints as deprecated, with the entries last updated in July and August 2026, before OpenAI's date. A model can stop being offered through your aggregator while the provider still runs it.
Platform APIs and MCP
Higher-level: not a generation endpoint but a workflow, with storage, chaining, and often a UI attached. Our own MCP server sits here, letting an agent generate images, video and music on 8frame credits without you building the pipeline.
What it exposes, from the connector's own configuration: eight image models plus editing, background removal and image upscaling, nine video tools (Kling, Veo, Wan, Seedance, Gemini Omni Flash, Hailuo, Grok Imagine, Pruna and Creatify Boreal) and music. Results come back as file links, not into a canvas. A generation still running after 10 minutes hands back a task ID to check later, and connector requests have a per-user daily credit cap as a spending safeguard. Defaults differ from the canvas: Veo through the connector starts on Veo 3.1 Standard (448 credits for 8 seconds with sound; the canvas starts on Fast, 168), and Kling there is Kling v3 Standard, image-to-video only, 86 credits for 5 seconds.
Where this is the wrong choice: if what you need is a raw endpoint inside your own product with your own queueing and storage, buy an aggregator or go direct. A workflow platform is overhead you will fight. It is also not the cheap route for a single model: an 8-second Veo 3.1 Standard clip is $3.20 on Google's API ($0.40 a second) and 448 credits on 8frame, about $4.48 at pack rate.
What the rates look like
Per second of generated video, from each provider's own pricing page, checked 2026-10-01. Verify before you budget; these move.
| Model (provider) | Price per second |
|---|---|
| Wan 3.0 (Alibaba Cloud, Singapore list price) | $0.05 / $0.10 / $0.20 at 480p / 720p / 1080p |
| Wan 3.0 Prime, the faster tier (Alibaba Cloud) | $0.068 / $0.14 / $0.28 |
| Grok Imagine Video (xAI) | $0.05 listed; xAI says resolution also affects the cost |
| Grok Imagine Video 1.5 (xAI) | $0.08 listed |
| Veo 3.1 Lite (Google) | $0.05 at 720p, $0.08 at 1080p |
| Runway Gen-4 Turbo (Runway) | $0.05 (5 credits) |
| Gemini Omni Flash (Google) | about $0.10 at 720p (billed $17.50 per 1M output tokens) |
| Veo 3.1 Fast (Google) | $0.10 at 720p, $0.12 at 1080p, $0.30 at 4K |
| Runway Gen-4.5 (Runway) | $0.12 (12 credits) |
| Veo 3.1 Standard (Google) | $0.40 at 720p and 1080p, $0.60 at 4K |
Alibaba's pages show list prices and say they exclude limited-time promotions. Google lists Veo prices as "video with audio price (default)".
Two things to note when budgeting. Audio changes the price on some routes: Runway's API bills Veo 3.1 at 40 credits a second with audio and 20 without, exactly double, while Google's own page gives only the with-audio price. And Gemini Omni Flash bills by output tokens, so Google's per-second figure is a conversion rather than the billing unit.
The lock-in questions to ask before integrating
What is the deprecation policy? OpenAI gave six months for Sora (notice 2026-03-24, shutdown 2026-09-24). Not everyone will.
Are model versions pinnable? If your product's output changes silently when a provider updates a model, you have a support problem you cannot debug.
What happens to in-flight jobs and stored outputs at shutdown? Rarely documented, and it decides whether a migration is a sprint or an incident. Sora's own guide said download links were valid for a maximum of one hour after generation, so only files you had copied out survived.
Can you dual-route? The cheapest insurance is a second provider integrated but not primary. With one in place, a deprecation is a configuration change rather than a project.
Migration checklist, if you were on Sora
Map every call site, including anything parsing the Videos API response shape or assuming Sora's fixed durations (OpenAI's last guide lists generations of up to 20 seconds). Test parity on two vendors, not one. Re-check audio assumptions, since Sora 2 returned clips with sound and some routes price audio separately. Then budget: the models in the table above run from $0.05 to $0.40 a second at 1080p or below, so the per-second spread is wide and the model you pick decides your total.
Full per-use-case migration table in Sora 2 API shuts down September 24. If you also need to know what happened to existing Sora content, see what happens to your Sora videos after Sept 24.
FAQ
What is the best AI video API for a startup? An aggregator, almost always. One integration buys you the ability to change models without a migration, which at startup stage is worth more than a few cents a second. Compare its per-model prices with the provider's own page before you commit.
Does 8frame have a video generation API? Not a raw one. We expose an MCP server so agents and tools like Claude can generate images, video and music on your 8frame credits, with results returned as file links. That suits workflow integrations and does not suit teams building their own pipeline.
How do I avoid another Sora? Integrate two providers and keep the second warm, and copy every output into your own storage the moment it is generated. Everything else is optimism.
Sources
- Sora shutdown date and affected models: OpenAI API deprecations, checked 2026-10-01
- Sora shutdown notice, durations and download-link validity: OpenAI video generation guide (archived by OpenAI), checked 2026-10-01
- Veo 3.1 and Gemini Omni Flash prices: Gemini API pricing (page updated 2026-10-01), checked 2026-10-01
- Wan 3.0 prices: Alibaba Cloud Model Studio, wan3.0-video and wan3.0-video-prime, checked 2026-10-01
- Grok Imagine Video prices: xAI models and pricing and xAI video generation guide, checked 2026-10-01
- Runway API rates, including third-party models: Runway API pricing, checked 2026-10-01
- fal Sora 2 endpoint status: fal model API, query "sora", checked 2026-10-01
- 8frame connector tools, defaults, wait limit and prices: 8frame MCP server configuration, read 2026-10-01; 8frame Veo prices: 8frame tool registry, read 2026-10-01
Pick the supplier relationship before the model, because the model will change. See how the models compare in best AI video generator 2026.
Try it without signing up: the 8frame canvas opens on a real board. The canvas is free and unlimited; generation is paid from $19/month.