[OpenAI](/entry/notebooklm-vs-chatgpt-projects-vs-claude-projects-which-for-research-2026) pulled the plug on Sora. The consumer app went dark on April 26, 2026; the API follows on September 24, 2026. Video inference is brutally expensive to run, engagement fell off after the novelty wore off, and the likeness-rights questions never got resolved. Whatever you were building on Sora, you're rebuilding now.
The three models worth rebuilding on:
| Price | Notes | |
|---|---|---|
| Veo 3.1 (Google) | Fast $0.09/sec, Standard $0.18/sec | native audio, best prompt adherence, via Gemini API / Flow |
| Kling 3.0 Pro | ~$0.095/sec | strongest motion realism, generous credit packs |
| Seedance 1.5 Pro (ByteDance) | ~$0.047/sec | cheapest by a wide margin, fast |
The math: a hundred 10-second clips runs about $47 on Seedance, $95 on Kling Pro, and $90–180 on Veo depending on tier. Seedance is roughly half the price of the other two and fast, which makes it the default for volume work where "good enough" is good enough. Veo 3.1 wins on quality-per-dollar once you factor in that it generates synced audio in the same pass — no separate sound design step — and its prompt following is the most reliable of the three. Kling sits in between, and its motion physics still look the most natural on complex movement.
Pick Veo if output quality and native audio matter and you're already in the Google ecosystem. Pick Kling for the best-looking motion and if you'd rather buy credit packs than meter per second. Pick Seedance when you're generating a lot and cost is the constraint. [Runway](/entry/runway-vs-pika-vs-luma-ai-video-generation-pricing-2026), [Pika](/entry/runway-vs-pika-vs-luma-ai-video-generation-pricing-2026) and [Luma](/entry/runway-vs-pika-vs-luma-ai-video-generation-pricing-2026) are all still around too if you want a full editing suite wrapped around the model rather than raw API access.