Seedance, Seedream, and OmniHuman sit under the Jimeng and Dreamina apps, the CapCut editor, the Pippit marketing agent, and the Volcano Engine cloud. Here is the whole ecosystem, mapped — and where the gaps still are.
2026-10-21 · by Moe Ameen
Through 2026, ByteDance quietly assembled something no other company has: a top-to-bottom AI video stack it owns at every layer. At the bottom sit the foundation models. Seedance is the text- and image-to-video family — Seedance 1.0 and 1.0 Pro arrived in June 2025, Seedance 2.0 in February 2026, and Seedance 2.5 on July 31, 2026 with native 30-second single-pass generation. Seedream is the sibling image model (Seedream 5.0 Pro was announced by the ByteDance Seed team on July 8, 2026). OmniHuman is the digital-human model that turns a single photo plus an audio track into a full-body talking video. These are not three teams renting each other's work; they are one in-house research program.
On top of the models sit the apps, which is where the naming gets confusing. Jimeng is the China-domestic creation app; Dreamina is its international counterpart, reachable with an English interface at dreamina.capcut.com. CapCut is the editor most Western creators already know, which now surfaces the models as "Dreamina Seedance." Pippit is the marketing-focused agent that grew out of CapCut Commerce Pro and targets small businesses and social sellers. All four run the same underlying Seedance and Seedream engines; they differ in region, audience, and workflow, not in the model inside.
The top layer is distribution to businesses: Volcano Engine (and its international arm, BytePlus) ships the models as a cloud API for studios and developers. ByteDance has pointed the stack at premium filmmaking and advertising, and the models reach a huge audience through tools people already use rather than only as a raw API.
Two caveats matter for creators reading this from outside China. First, availability is uneven: the consumer rollouts of newer models have launched first in a set of non-US markets, and US access to several surfaces has been functional but uncertain rather than guaranteed. Second, the legal picture is live — in August 2026 ByteDance and the Motion Picture Association signed a memorandum of understanding on IP protection for Seedance and Seedream, but copyright pressure from studios is ongoing. Treat specific regional availability, caps, and pricing as a snapshot that shifts.
The smart way to use an ecosystem this big is to not get trapped in it. Each ByteDance app is a walled garden: Dreamina makes a clip, CapCut edits it, Pippit wraps it for a product — and then each one stops at an export file, inside ByteDance's world, with your audience still waiting on eight other platforms. Kompozy is the deliberately vendor-neutral layer that sits downstream of all of it. Export a Seedance clip from Dreamina, an OmniHuman avatar, a CapCut edit, or nothing from ByteDance at all, bring it into Kompozy, and it becomes finished posts: branded captions for silent feeds, reframing to 9:16, 1:1, or 16:9, a hook overlay through HyperFrames on the first second, then scheduling and publishing across Instagram, Facebook, TikTok, YouTube, LinkedIn, X, Pinterest, and Threads — plus blog and email — from one queue. It does not care which model you used, which is exactly the point: if US access to a given surface wobbles, your publishing pipeline does not.
Kompozy also closes the ecosystem's real gap, which is net-new on-brand output ByteDance's apps do not finish for you. It generates HeyGen talking-head Persona Shorts and avatar video composited into brand-exact templates, multi-slide carousels, face-locked persona images, quote graphics, blog articles, and newsletters — all governed by a single Persona Brief so your voice and identity stay consistent whether the footage came from Seedance, from HeyGen, or from your own camera. Use ByteDance's stack to make striking scenes; use Kompozy as the one place those scenes — and everything around them — turn into a scheduled, on-brand week of content across every platform your audience uses.
ByteDance runs a full stack. The foundation models are Seedance (text/image-to-video), Seedream (image), and OmniHuman (a photo-plus-audio digital-human model). On top of them sit the consumer apps: Jimeng (China), Dreamina (international), the CapCut editor, and the Pippit marketing agent. For businesses, the models ship as a cloud API through Volcano Engine and its international arm, BytePlus.
They are four surfaces onto the same underlying Seedance and Seedream engines. Jimeng is the China-domestic creation app and Dreamina is its international counterpart; CapCut is the general video editor that surfaces the models as "Dreamina Seedance"; and Pippit is a marketing-focused agent for small businesses and social sellers. The model is the same — region, audience, and workflow are what differ.
It depends on the surface and the week. Newer consumer rollouts have launched first in a set of non-US markets, and US access to several ByteDance video surfaces has been functional but uncertain rather than guaranteed. There is also ongoing copyright pressure from studios, despite an August 2026 memorandum of understanding between ByteDance and the Motion Picture Association. Confirm current availability on the official apps before relying on any one of them.
The ByteDance apps generate and edit, but each stops at an export file inside its own world. Bring that clip into Kompozy to add branded captions, reframe per platform, stack a hook overlay, and schedule and publish across Instagram, TikTok, YouTube, LinkedIn, X, and more from one queue — plus generate the persona video, carousels, blogs, and newsletters the ByteDance apps do not finish for you. Kompozy is vendor-neutral, so it works the same no matter which ByteDance tool the footage came from.