Sora is here

## Sora is live — what actually matters

### 1. Context: two years of anticipation, a constrained launch

Sora was announced in February 2024 with striking demos — cinematic shots, multi-second temporal coherence, convincing simulated physics. Ten months later, the product ships on sora.com with meaningful constraints: 20-second maximum, 1080p ceiling, and access initially gated to ChatGPT Plus and Pro subscribers. This is not the model from the February demos — OpenAI confirmed the public version differs from internal previews, with stricter content guardrails and reduced capability on complex scenes.

Before this launch, the AI video generation market was led by Runway (Gen-3 Alpha, up to 10 seconds at 768p), Kling (Kuaishou, up to 2 minutes but uneven quality), Pika Labs (v1.5, 3-5 seconds), and Luma Dream Machine. No Western consumer-facing player had reached native 1080p with 20 seconds of temporal coherence.

### 2. The technical specs that matter

**Resolution and duration**: 1080p / 20 seconds is a real quantitative step. Runway Gen-3 caps at 10 seconds. Kling reaches longer durations but with visible degradation on complex motion. Sora's signal-to-noise ratio across 20 seconds still needs real-world validation — controlled demos don't reflect production use cases.

**Aspect ratios**: landscape (16:9), portrait (9:16), square (1:1). Portrait support is critical for social content creators (TikTok, Reels, Shorts) — this is the first tier-1 model to support it natively without post-generation cropping.

**Asset remixing and blending**: the least-discussed feature but potentially the most impactful for professional workflows. Bringing your own assets (images, video clips) to extend or blend opens post-production use cases that pure text-to-video models don't cover. Runway had partial capability with Gen-2; Sora positions it as a core feature.

**Pricing**: ChatGPT Plus ($20/month) grants access with generation limits not yet precisely documented. ChatGPT Pro ($200/month) offers priority access and higher quotas. No public API announced at launch — which limits integration into third-party pipelines.

### 3. The likely losers

**Runway** is most exposed. Gen-3 Alpha was the Western qualitative benchmark since June 2024. With Sora at 1080p/20s embedded in the OpenAI/ChatGPT ecosystem, Runway loses its distribution advantage and must accelerate Gen-4 (no date announced). Its freemium credit model is also under pressure against Sora being bundled into existing subscriptions.

**Pika Labs** and **Luma** are less directly threatened short-term — they target different segments (Pika on fast editing, Luma on character consistency) but lose narrative ground.

**Adobe** (Firefly Video, in beta) sees its launch window shrinking. Firefly Video promised Creative Cloud integration and clear commercial usage rights — two advantages Sora doesn't yet explicitly guarantee.

**Low-end video production shops** (explainer videos, simple ads, repetitive social content) are the first to feel economic pressure. 20 seconds at 1080p covers exactly the standard social ad format.

### 4. What remains unclear — and what's missing

The absence of an API at launch is a meaningful signal: OpenAI controls the access point and avoids immediate commoditization. This slows enterprise adoption but protects margins.

Commercial usage terms haven't been clarified publicly with the same precision as DALL-E 3. For agencies and brands, this is a real blocker.

Character consistency across multiple generations — the holy grail of AI video production — is not documented as a feature of public Sora. Kling and certain Runway/ComfyUI workflows remain superior on this point.

Finally, generation latency (wall-clock time for 20 seconds at 1080p) is not communicated. In production, a 5+ minute generation time per clip fundamentally changes the economic calculus.

Sora is a solid launch, not a finished product. The real competition starts when the API ships.