AI Video Generation Tools
Landscape of AI video generation (text-to-video, image-to-video, editing, native audio), as of 2026-10-02. Vendor notes carry the detail; this note is the map.
Current models
| Model | Vendor | Status / notes |
|---|---|---|
| google-veo (Veo 3.1, Lite; Gemini Omni Flash) | Gemini API (2026-10-02): Veo 3.1 Standard, Fast and Lite tiers (per-second pricing per the docs), all marked preview as of 2026-10-02. Gemini Omni Flash is the docs’ newer default (token-priced). Flow app: google-flow | |
| kling | Kuaishou | Kling 3.0 (announced 2026-02-05); Kling 4.0 reported as announced late Sept 2026 (30 s, 4K; release timing not verified; Wikipedia, secondary) |
| seedance | ByteDance | Seedance 2.5 (July 2026, per Wikipedia, secondary), up to 30 s |
| wan-2-dot-5 | Alibaba | Wan 3.0 latest (API); Wan 2.2 open, Apache 2.0 |
| minimax-hailuo-video | MiniMax | H3 (2026-07-31 per MiniMax blog), 15 s at 2K; open weights reportedly announced |
| runway | Runway | Gen-4.5, world models (GWM Worlds 2, 2026-09-03) |
| FLUX 3 Video | Black Forest Labs | listed on BFL blog 2026-08-04; see flux-dot-1 |
| heygen | HeyGen | avatar/presenter video (builds on third-party models) |
| openai-sora | OpenAI | Discontinued: app ended 2026-04-26; the Sora 2 API was scheduled for removal on 2026-09-24 and secondary reports say it was switched off that day (not independently observed) |
Access paths
Vendor apps and APIs; aggregators fal-dot-ai (per-second pricing across video models; not re-checked), kie-dot-ai, and local workflows in comfyui for open weights (Wan 2.x, LTX, Hunyuan).
Trends
Editorial observations (opinion, not systematically verified):
- Native synchronised audio appears in several current models.
- Clip lengths are growing (15 s to 30 s in the vendor listings above).
- Some video models are being framed as “world models” (Runway GWM, Google Omni).
- Legal friction has been reported: studio cease-and-desist letters over Seedance 2.0 (not verified here); Midjourney suits. See midjourney.
Choosing
Suggestions (opinion): a vendor API for reliability, an aggregator for price shopping (verify the listed model still works; Kie.ai’s docs still list Sora), and open weights for privacy. Always check likeness, consent and licensing before commercial use.
Open items
- Comparative rankings (blog roundups still name Seedance 2.0 / Kling 3.0 / Veo 3.1 / Runway Gen-4.5 as the leaders) are editorial opinions and are not repeated as facts.
- Per-second prices were not re-verified and are not recorded here; the old note’s Kie.ai price and “Sora 2 free with ChatGPT Pro” claims stay dropped.
- Seedance 2.5 / Wan 3.0 / H3 details live in their own notes, which have their own verification status.
Sources
- https://ai.google.dev/gemini-api/docs/models (accessed 2026-09-30)
- https://en.wikipedia.org/wiki/Seedance , https://en.wikipedia.org/wiki/Kling_AI (accessed 2026-09-30)
- https://www.alibabacloud.com/help/en/model-studio/video-generation (accessed 2026-09-30)
- https://www.minimax.io/blog/minimax-h3 (accessed 2026-09-30)
- https://runway.com/research (accessed 2026-09-30)
- https://fal.ai/pricing (accessed 2026-09-30)
- https://docs.kie.ai/ (accessed 2026-09-30)
- https://ai.google.dev/gemini-api/docs/pricing (accessed 2026-10-02)
- https://en.wikipedia.org/wiki/Sora_(text-to-video_model) and https://help.apiyi.com/en/sora-2-api-shutdown-alternatives-2026-en.html (Sora API shutdown, secondary; accessed 2026-10-02)