AI video generator that turned text prompts and images into short cinematic clips — its consumer app is now shut down.
Last verified · 2026-06-26 · by Moe Ameen
Haiper AI was an AI video generation platform that turned a text prompt or a still image into a short cinematic clip. It was founded in 2021 in London by Yishu Miao (CEO) and Ziyu Wang (CTO), two researchers who had previously worked at Google DeepMind. The company stayed in stealth for two years, then emerged in March 2024 with a $13.8M seed round led by Octopus Ventures (with 5Y Capital), pitching an "all-in-one visual foundation model" for studios, publishers, and individual creators.
Its most capable public release was Haiper 2.0, launched on 28 October 2024. That version covered text-to-image, text-to-video, image-to-video, and video extension, generated clips up to about six seconds long, and added templates for things like logo and product animations, an HD upscaler, and keyframe conditioning. A 2.5 model and a public API followed on 18 December 2024, alongside a partnership that put Haiper's model inside the VEED online editor — a clear signal the company was shifting from its own consumer app toward licensing the model into other products.
The big thing to know in 2026: Haiper's consumer product is gone. The company shut down its free, browser-based video app in February 2025 without warning. In March 2025, cofounders Yishu Miao and Ziyu Wang — along with machine-learning researcher Edward Hayes — joined Microsoft AI under Nando de Freitas to work on multimodal generative systems; reporting at the time framed it as a talent move rather than an acquisition of the company. By mid-2025 Haiper's video generation models were sold to NetMind.AI, a London-based decentralized AI-compute platform that folded the technology into its B2B inference infrastructure. So the model survives inside an enterprise API, but the Haiper.ai you could sign up for and generate clips on is not coming back.
Because the consumer surface is discontinued, treat any "how to use Haiper" tutorial that predates 2025 as a historical record, not current instructions. If you arrived here looking for a live text-to-video tool, you'll need a different generator — the practical question is which engine you build your actual publishing workflow on so a single model's shutdown can't strand you again.
Haiper's story is the argument for not building your content workflow on a single consumer video app. The model was good and the company still shut its doors to creators overnight — anyone who had a Haiper-based posting routine lost it in February 2025. Kompozy is the durable layer underneath that volatility: it generates video itself and never depends on one provider's consumer product staying alive. Where Haiper rendered a clip and stopped, Kompozy produces finished, captioned, correctly-sized posts and publishes them across nine platforms.
Concretely: a short AI-generated clip — whether it came from a Haiper-style text-to-video tool or from one Kompozy makes natively — drops into Kompozy and comes out as a finished short. Kompozy burns in branded, on-style captions, reframes the clip to each destination's aspect ratio, and lets you stack hook text and overlays through HyperFrames so the silent-autoplay first second actually lands. Then it schedules and publishes the same short to TikTok, Reels, YouTube Shorts, X, LinkedIn, and the rest in one pass. And because Kompozy generates net-new video on its own — Persona Shorts (talking-head avatar video), Clipped Shorts from long-form, Listicle Videos over stock footage, Marketing Shorts — you are not waiting on any one model to come back online to keep posting. One source idea fans out into a week of cross-platform content, all written in your voice through the Persona Brief, instead of a single orphaned six-second render.
No. Haiper shut down its consumer video app in February 2025 without warning. The Haiper.ai sign-up product with its free tier is not coming back. The underlying models were sold to NetMind.AI in mid-2025 and now live inside that company's B2B inference API, not a consumer tool you can use directly.
Haiper closed its consumer business in February 2025. In March 2025 its cofounders Yishu Miao and Ziyu Wang, plus ML researcher Edward Hayes, joined Microsoft AI to work on multimodal generative systems — widely reported as a talent move rather than a company acquisition. By mid-2025 Haiper's video models were sold to NetMind.AI, which integrated them into its enterprise platform.
Haiper was founded in 2021 in London by Yishu Miao (CEO) and Ziyu Wang (CTO), both former Google DeepMind researchers. It emerged from stealth in March 2024 with a $13.8M seed round led by Octopus Ventures, with participation from 5Y Capital.
Before it shut down, Haiper generated short clips (up to about six seconds) from text prompts or a still image, plus image-to-video animations, video extensions, template-driven motion like logo reveals, and AI still images. Its most capable public release was Haiper 2.0 in October 2024, followed by a 2.5 model and an API in December 2024.
For a live text-to-video model you would pick a current generator. But if your goal is publishing consistently, the more durable move is a content engine like Kompozy that generates video itself (talking-head Persona Shorts, clipped shorts, listicle videos) and publishes across nine platforms — so a single model going offline never breaks your posting schedule.