The AI video market after Sora's exit — with real pricing: $0.50 to $2.50 per 10-second clip, who leads the rankings now, and how to choose per shot.
| You need… | Pick | Why |
|---|---|---|
| Cinematic single shots with sound | Google Veo 3.1 | 4K + synchronized 48 kHz audio, the quality bar |
| Best quality per dollar | Kling 3.0 | Native 4K/60fps, roughly a fifth of Veo's per-clip cost |
| Longest single generation + lip-sync | Seedance 2.0 | Up to 15s per call, sub-120ms lip-sync, 12-file input |
| A production pipeline, not just clips | Runway Gen-4.5 | Keyframes, Video-to-Video, motion brush |
| Your Midjourney image, moving | Midjourney video | 5–21s, unbeatable if the source aesthetic is already yours |
OpenAI pulled Sora on March 25, 2026; its API sunsets September 24. The model that ignited the text-to-video race exited it — the first full retreat from a headline modality by a frontier lab (our OpenAI tracker covers why).
The reported economics were not close: roughly $15M/day in operating cost against ~$2.1M in total lifetime revenue, at about $1.30 per 10-second clip in raw generation cost. Sora didn't lose a quality race. Its unit economics never closed.
The 2026 Artificial Analysis leaderboard is not what late-2025 coverage predicted:
Leaderboard position measures single-clip output quality on a fixed prompt set. It does not measure whether the tool helps you finish a project. Runway's fall on the index coincided with it remaining one of the most-used tools in actual production teams — because workflow value and clip-quality value are different products.
Two of the top three slots and four of the top ten going to ByteDance, Alibaba and Kuaishou is the structural change of 2026. It's also why the price floor collapsed: these labs compete on cost per second in a way the Western incumbents priced against reluctantly.
Google's flagship remains the bar for cinematic output: 4K resolution, synchronized 48 kHz audio generated with the video, and the most realistic camera movement in blind comparisons. Delivered through Google Flow with the strongest prompt adherence of the majors.
| Spec | Veo 3.1 |
|---|---|
| Resolution | 4K |
| Clip length | ~8 seconds per generation |
| Audio | Native, synchronized, 48 kHz |
| API price | $0.50/s video, $0.75/s with audio (Vertex AI) |
| Per 10s clip | ~$2.50 |
Less granular mid-generation control than Runway — you're renting a camera, not an edit suite. Shorter maximum clip than Seedance. And it is decisively the most expensive option per second, which matters the moment your project needs more than a handful of hero shots.
Hero shots, advertising, anything where single-clip quality is the deliverable and the clip count is small. It also anchors Gemini's broader multimodal push.
Native 4K at 60 fps plus a genuinely generous free tier, at roughly $0.50 per 10-second clip — about a fifth of Veo's cost for the same duration. Per-second API rates through providers like fal.ai run near $0.029/s, the cheapest of the majors.
Prompt adherence lags on complex multi-subject scenes. Generated audio quality sits clearly below Veo 3.1 and Seedance 2.0. The integration and ecosystem story is thinner than Google's or Runway's.
High-volume workflows where cost per usable clip dominates: storyboards, drafts, social volume, pre-visualization. Four top-10 entries means the quality is not a compromise — the constraint is control, not fidelity.
Up to 15 seconds in a single call — nearly double Veo's practical ceiling — with 48 kHz native audio and sub-120 ms lip-sync alignment. If your shot involves someone speaking, this is the shortest path to something usable.
Seedance accepts up to twelve reference files in one generation. That's a different interaction model from "prompt plus one image," and it's why it leads on creative flexibility: you can specify subject, style, environment and motion reference simultaneously rather than sequentially.
Around $0.60 per 10-second clip — close to Kling, far below Veo. Combined with the #1 leaderboard position, it's the current default recommendation for anyone without an existing pipeline commitment.
Runway optimizes for a different prize: the workflow. Gen-4.5 carries the full control suite — Image-to-Video, Keyframes (pin start and end states), Video-to-Video, motion brush, camera controls, style references. It is where generations become finished work rather than lucky pulls.
Raw single-shot fidelity now trails all three leaders, and there's no native audio. The index drop is real and shouldn't be explained away.
Teams shipping video weekly, iterative shot development, and Sora-workflow refugees — Keyframes plus Video-to-Video covers most "extend or vary this shot" patterns that Sora users relied on. Deeper dive in the Runway tracker and the hands-on production guide.
| Model | Approx. cost |
|---|---|
| Kling 3.0 | ~$0.50 |
| Seedance 2.0 | ~$0.60 |
| Sora 2 (sunsetting) | ~$1.00 |
| Veo 3.1 | ~$2.50 |
You will see Kling quoted anywhere from $0.029/s to $0.07/s and Veo from $0.05/s to $0.75/s. Both ranges are real — they reflect different tiers (Lite vs full), different providers, and audio on or off. Per-clip figures are the more stable comparison; always confirm the exact tier before budgeting.
At a 5× spread between cheapest and most expensive for comparable duration, "which model is best" stops being the operative question for volume work. Cost per usable second takes over — and usable means after your rejection rate, which varies more by model than the sticker price does.
Sora's exit proved that scale doesn't guarantee survival in this category. This tier is one funding winter away from feeding the leaders. Don't build a pipeline that only one specialist tool can run.
| Deliverable | Model |
|---|---|
| One hero shot with sound | Veo 3.1 |
| Someone speaking on camera | Seedance 2.0 |
| 50 social clips this month | Kling 3.0 |
| A sequence you'll iterate for a week | Runway Gen-4.5 |
| Motion from images you already made | Midjourney video |
Choose per shot, not per project. The per-clip price spread and the capability differences are both large enough that a single-vendor project is almost always leaving quality or money on the table. This is the single biggest workflow change between 2025 and 2026.
Sources: LaoZhang AI — Seedance 2.0 vs Kling 3.0 vs Sora 2 vs Veo 3.1 · BuildFastWithAI — best AI video July 2026 · Unifically — Veo 3.1 vs Seedance API pricing · TokenMix — AI video API cost survey · Digital Applied — the market after Sora · Teamday — best AI video models 2026
Last updated: July 30, 2026
Seedance 2.5 brings 30-second audio-video generation, 50 mixed references and timed editing. Here is what is official, priced and still worth testing.
MiniMax H3 generates 2K video with native stereo audio, publishes per-second pricing, and offers H3-Base weights under a Community License.
Hermes Agent blends persistent memory with self-improving skills, but its security and outcome depend on the backend, approvals, and review you choose.
OpenClaw is a high-capability agent runtime whose value depends on deliberate gateway security, skill review, and operating discipline.
AI music in 2026: Suno v5.5's capabilities and legal defiance, Udio's settlement path, the Warner deal retiring unlicensed models, and creator rules.
The honest 2026 free-tier comparison: Gemini's generous bundle, Claude's quality-first plan, ChatGPT's capped breadth, and when free stops being enough.
Archived review of GPT-5's August 2025 launch: the unified-router design, the 4o backlash, what improved — and how it reads from mid-2026.
An honest 2026 assessment: what GPT-5.5, Dreaming V3 memory and agents actually deliver, where Claude and Gemini beat it, and who should pay for Plus.
