- - 🎬 Text-to-video endpoint of MiniMax's H3 generation video model.
- - 🆕 MiniMax describes H3 as a new-generation open multimodal video model.
- - 🏢 Built by MiniMax, a Chinese lab spanning text, audio and video.
- - 🔧 Sibling endpoints cover image-to-video and reference-to-video conditioning.
- - 📅 Released July 2026 alongside the rest of the H3 family.
- - 💬 Prompt-only workflow: describe a shot, receive a generated clip.
- - 🌐 Available through hosted APIs rather than local weights on this catalog.
MiniMax is an AI company building generative models across multiple modalities, with a focus that spans both language understanding and audio creation. Their rapid release cadence in early 2026—delivering several new models within just a few months—reflects an ambitious and…
Explore 10 more models by Minimax →MiniMax H3 (text-to-video) is the prompt-driven entry point to MiniMax's H3 video generation family, released in July 2026. Where earlier MiniMax video systems were positioned as standalone clip generators, MiniMax's own developer documentation frames H3 as a "new-generation open general-purpose multimodal video model," signalling a shift toward one model handling several conditioning modes rather than a set of separate task-specific checkpoints.
This variant takes a written description and synthesises a video clip with no visual input required, making it the most open-ended of the three H3 surfaces. Its siblings share the same underlying generation stack but change the conditioning: MiniMax H3 animates a supplied still frame, while MiniMax H3 R2V uses reference material to carry subject or style identity into the output. Choosing between them is mostly a question of how much control you want over the starting appearance.
Because detailed technical specifications for H3 are not published in the primary documentation available here, this page avoids quoting resolution, duration or benchmark figures from secondary sources. What is verifiable is the family framing and release timing.
H3 sits within a broader MiniMax lineup that includes the multimodal text model MiniMax M3 Preview, the MiniMax Speech-02 HD voice system, and MiniMax Music 2.6 — components that can be combined into an end-to-end script, narration and video pipeline.
This About section is AI-generated from public sources via VeniceStats + Venice inference, with no human editing. It may contain inaccuracies.
Daily usage · MiniMax H3 · Organic sellers
Top sellers · MiniMax H3 · Organic
- Open Forge59.6%
- Venice.ai Proxy37.0%
- antseed-tidal-falcon-88cd1.9%
- attacompute.ai1.0%
- ▲ Apex Ant0.5%
| Seller | Health | Rep↓ | Sales | Video | Routing | $ / s | $ / video |
|---|---|---|---|---|---|---|---|
| Open Forge 0x1d90…b0aa | healthy | 91.11 | $102.05 | 33.8 min | #1 | $0.05 | — |
| ▲ Apex Ant 0x73b4…e736 | — | 81.71 | $0.90 | 20 s | #3 | $0.045 | — |
| attacompute.ai 0xd5e7…7be0 | — | 51.22 | $1.77 | 15 s | gated | $0.116 | — |
| antseed-tidal-falcon-88cd 0x80d1…88cd | — | 12.78 | $3.17 | 30 s | gated | $0.011 | — |
| OpenCheap 0x0ee1…6c41 | — | 12.43 | — | — | gated | $0.057 | — |
| Venice.ai Proxy 0x1f22…18c9 | failingupstream error · 3.7 h · stale | 86.84 | $63.25 | 20.6 min | #2 | $0.055 | — |
"Best price" and the seller table are live Antseed catalog data (advertised $/1M tokens — or $ per generated image for unit-billed image models — not settled amounts). Reputation = buyer trust score (0-100, the Antseed SDK's own formula). "Routing" = the SDK's default buyer routing (what the VPR desktop app ships with): a trust ≥ 60 gate on the effective reputation, then cheapest-first among routable sellers; live failover state (per-peer cooldowns) is buyer-side runtime and not included. Model knowledge (TLDR, provider, About) via the VeniceStats enrichment layer. Advertised catalog, not the model used in any specific purchase. "Usage on Antseed" counts only settlements whose buyers share the per-model split on-chain (metadata v2/v3, opt-in), so every usage figure is a lower bound.