Cloud platforms dominate AI video. Runway, Google Veo, Kling, Hailuo, Luma, and Pika all generate short clips from text or an image in the browser, priced in monthly credits.
OpenAI's Sora, the best-known name in AI video, was discontinued in April 2026. It appears in the table for completeness, not as a current option.
Running video models on your own computer became realistic in the last year. Wan 2.2's small variant generates 720p on a single RTX 4090, HunyuanVideo 1.5 runs from about 14 GB of VRAM, and CogVideoX runs on modest GPUs. Expect minutes of compute per few seconds of clip.
Snapshot of 7 June 2026. 21 entries in this group, about 18 with a free option and 5 that can run on your own hardware. Prices and versions change often, and each name links to its official page. The full directory has search and filters across all categories.
How to Choose
Try it free:
Pika's free tier includes 80 monthly credits at 480p with commercial use allowed. Kling refreshes free credits daily. Runway gives a one-time batch of 125 credits.
Sound in the video:
Google Veo, Grok Imagine, Seedance, and Vidu generate audio natively with the picture. LTX-2 is the open-weight option with synchronized audio.
Talking-head videos:
Hedra, HeyGen, and Synthesia animate a person from a photo plus audio. A different job than open-world scene generation.
Run it at home:
Wan 2.2 (the 5B variant) on a single RTX 4090, HunyuanVideo 1.5 from about 14 GB of VRAM, CogVideoX from about 5 GB. All open-weight.
Professional pipeline:
Runway pairs generation with editing tools and performance capture, and resells Veo and Kling inside one workspace.
All 21 Video generation Entries
| Tool | Maker | What it is | Cost | Where it runs | Hardware |
|---|---|---|---|---|---|
| Adobe Firefly Video | Adobe | Generate, translate, and lip-sync video. Positioned as commercially safe. | Free + paid | Cloud | Any device |
| CogVideoX | Zhipu / THUDM | Early, widely-adopted open video family with modest hardware needs. | Open-weight | Local | From ~5GB VRAM |
| Google Veo | Google flagship video model with native synchronized audio. | Free + paid | Cloud | Any device | |
| Grok Imagine (Video) | xAI | xAI native video-plus-audio model inside Grok. | Free + paid | Cloud | Any device |
| Hailuo | MiniMax | Video generator known for strong prompt adherence and complex motion. | Free + paid | Cloud | Any device |
| Hedra | Hedra | Animates a single photo into a talking character driven by audio. | Free + paid | Cloud | Any device |
| HeyGen | HeyGen | AI avatar video for marketing and dubbing across 175+ languages. | Free + paid | Cloud | Any device |
| Higgsfield | Higgsfield AI | Creative platform running many models with camera-motion presets. | Free + paid | Cloud | Any device |
| HunyuanVideo 1.5 | Tencent | Lightweight open video model. Note the territorial license limits. | Open-weight | Cloud or local | ~14GB VRAM and up |
| Kaiber | Kaiber AI | Multi-model creative video studio with a timeline editor. | Free + paid | Cloud | Any device |
| Kling | Kuaishou | High-quality Chinese video model with strong motion. 4K on higher tiers. | Free + paid | Cloud | Any device |
| LTX-2 | Lightricks | Open audio-plus-video model with native 4K and synchronized audio. | Open-weight | Cloud or local | Consumer GPUs |
| Luma Dream Machine | Luma AI | Text and image-to-video with an agentic creative workflow. | Free + paid | Cloud | Any device |
| Mochi 1 | Genmo | Open diffusion model with strong motion and prompt adherence. 480p preview. | Open-weight | Cloud or local | ~22GB+ VRAM |
| OpenAI Sora | OpenAI | Formerly OpenAI flagship video model. The product has been retired. | Discontinued | Cloud | Any device |
| Pika | Pika Labs | Consumer video with editing effects. Up to 1080p. | Free + paid | Cloud | Any device |
| Runway | Runway | Flagship Western text and image-to-video platform with editing tools. | Free + paid | Cloud | Any device |
| Seedance | ByteDance | ByteDance video model with native audio and real-human video support. | Paid | Cloud | Any device |
| Synthesia | Synthesia | Enterprise AI avatar platform for training and corporate video at scale. | Paid | Cloud | Any device |
| Vidu | Shengshu | Chinese video model with audio sync and multi-shot generation. | Free + paid | Cloud | Any device |
| Wan 2.2 | Alibaba | Open Mixture-of-Experts video family. Text, image, and a fast 5B hybrid. | Open-weight | Cloud or local | 5B on RTX 4090. A14B ~80GB |
Common Questions
Is Sora still available?
No. OpenAI shut down the Sora web and mobile apps on April 26, 2026, and the API is scheduled to sunset on September 24, 2026.
Can I make AI video for free?
Yes, within credit limits. Pika gives 80 credits a month with commercial use allowed, Kling refreshes daily credits, Hailuo and Vidu have free tiers with watermarks or non-commercial terms, and Runway gives a one-time 125 credits.
Can my gaming PC generate video?
Yes, slowly. Wan 2.2's 5B model produces a 5-second 720p clip in about 9 minutes on an RTX 4090. HunyuanVideo 1.5 runs from about 14 GB of VRAM with offloading. CogVideoX runs on modest hardware at lower resolution.
Keep Reading
Verified articles related to this page:
Other model pages: