Agnes AI put a no-cost version of its 2.5 video model on the Pavo platform — 720p text- and image-to-video clips of a few seconds, on a daily credit allowance.
2026-08-27 · by Moe Ameen
Agnes AI, a Singapore startup, has released a free "Flash" tier of its Agnes Video 2.5 model on Pavo, its AI creation platform. Reported August 27, 2026, the free tier generates 720p clips of roughly 4 to 12 seconds from a text prompt or a reference image, on a daily credit allowance with a low requests-per-minute cap. It's the company's on-ramp to the 2.5 video model: a way to generate real AI video without paying.
The Flash tier sits under a paid Standard tier that removes the caps and is priced per minute of output through the API. Agnes AI positions Video 2.5 as faster than the previous 2.0 generation, with steadier camera motion and more plausible physics, and supporting text-to-video, image-to-video, and reference or keyframe-style inputs. On the Artificial Analysis blind-test leaderboard the 2.5 series reached an Elo around 1082, up from about 917 for the 2.0 generation — putting it in range of mainstream commercial video models.
Pavo itself bundles text, image, and video generation into one workspace with an "infinite canvas" and an Agent mode that can take a single instruction and run the flow from script to finished clip. Treat the free-tier caps, resolution, credit counts, per-minute pricing, and benchmark figures as launch-window details — Pavo's own pages are the source of truth as they settle.
The pattern worth noticing here is commoditization. Every few weeks another capable video model ships a free tier — Agnes Video 2.5 Flash is the latest — and the practical effect is that "can you generate a clip" stops being the thing that separates creators. When a cinematic shot is free and fast for everyone, the bottleneck moves to what happens after the render: captioning it, reframing it for each platform, keeping it on-brand, expanding the idea into other formats, and actually publishing it on a schedule. That downstream half is exactly what Kompozy does, and it's the half a clip model like Agnes never touches.
So the way to act on this launch is to treat the free tier as a supply of raw footage and let Kompozy finish it. Generate your hook or B-roll on Pavo, bring the clip into Kompozy, and it burns in captions, reframes to 9:16, 1:1, and 16:9, and — under a single Persona Brief — fans the same idea into captioned Persona and Clipped Shorts, an avatar-led Persona HeyGen segment, brand-exact carousels, quote graphics, a blog, and a newsletter, then schedules and publishes the set across the eight social platforms plus blog and email. Kompozy has no arbitrary text-to-video generator of its own, and it doesn't need one: cheap generation is now everywhere, and the finishing-and-distribution layer is where the leverage is.
Yes — Flash is the free tier of Agnes Video 2.5 on the Pavo platform. It generates 720p clips of roughly 4–12 seconds from a text prompt or reference image, on a daily credit allowance with a low requests-per-minute cap. A paid Standard tier removes the caps and is priced per minute of output. Confirm current limits on Pavo's pages.
Both come from Agnes AI, a Singapore startup founded in 2025. Pavo is its free AI creation platform combining text, image, and video generation with an infinite canvas and an Agent mode; Agnes Video 2.5 is its video model, with a free Flash tier and a paid Standard tier.
Agnes generates a raw, uncaptioned clip and publishes nowhere. Bring it into a content engine like Kompozy to caption it, reframe it to 9:16, 1:1, and 16:9, hold it to a brand voice, expand the idea into carousels, a blog, and a newsletter, and schedule it across the eight social platforms plus blog and email.