Text-to-Video Foundation Models

Large models that generate video clips from text or images.

Tier Grades

Tier A: Frontier AI labs

feasible

Training on billions of images and videos with a 62-author team is frontier-lab work.

Tier B: Indonesian AI labs

not feasible

Pretraining at that scale is not feasible for an Indonesian lab.

Tier C: Institution units such as ITB

stretch

A university unit can fine-tune the 1.3B model, which runs in 8.19 GB VRAM.

Tier D: Organisations such as SEMAI (PoT, PoC, PoV)

feasible

APIs and the small open model support marketing and training video pilots.

Where this comes from

Tiers: a) frontier AI labs, b) Indonesian AI labs, c) institution units such as ITB, d) organisations such as SEMAI delivering proof of technology, concept and value. Each rationale cites the anchors below.

Anchors

ComputeWan offers 1.3B and 14B models; the 1.3B model needs only 8.19 GB VRAM. Training compute is not stated in the abstract.
DataThe 14B model is trained on billions of images and videos.
Team size (proxy)Proxy only (author count of an exemplar paper, not a headcount): 62 authors on the Wan paper (arXiv 2503.20314).
Time horizonMonths.
Shelf-life signalsA new open model tops the video arenas, Copyright actions against video models
Pipeline roleTier A trains video models; Tier C fine-tunes small open ones; Tier D uses them in content pilots.

Where this comes from

Team size is a proxy (author counts of exemplar papers), not a headcount.

Who and What

Exemplars

Observed actors

Alibaba (Qwen) Google DeepMind

Google DeepMind Kuaishou

Kuaishou

Tencent

Linked models

Kling 3.0video generationexemplar

HunyuanVideo 1.5video generationexemplar

Wan 2.1video generationexemplar

Veo 3video generationexemplar

Vidu Q4 Previewvideo generationproduced by

MAGI-2 Previewvideo generationproduced by

Utopai Xvideo generationproduced by

LTX-2.5video generationproduced by

Wan 3.0video generationproduced by

Gemini Omni Flash 1.1video generationproduced by

MiniMax H3 Maxvideo generationproduced by Agnes-Video-2.5video generationproduced by

Seedance 2.5video generationproduced by

MiniMax H3 (Hailuo 3)video generationproduced by

FLUX 3 Videovideo generationproduced by

Muse Videovideo generationproduced by

HappyHorse 1.1video generationproduced by

Bernini-Rvideo generationproduced by

Gemini Omni Flashvideo generationproduced by

Grok Imagine Video 1.5video generationproduced by

HappyHorse 1.0video generationproduced by

Veo 3.1 Litevideo generationproduced by

PixVerse V6video generationproduced by

SkyReels V4video generationproduced by

+80

Seedance 2.0video generationproduced by

Kling 3.0 Omnivideo generationproduced by

Veo 3.1video generationproduced by

LTX-2video generationproduced by

Vidu Q3 Provideo generationproduced by

Pika 2.5video generationproduced by

Hailuo 2.3video generationproduced by

LongCat-Videovideo generationproduced by

Kling 2.5 Turbovideo generationproduced by

Wan 2.2video generationproduced by

Wan 2.2 A14Bvideo generationproduced by

Wan 2.2 TI2V-5Bvideo generationproduced by

Seedance 1.0video generationproduced by

Kling 2.1video generationproduced by

Vidu Q1video generationproduced by

SkyReels-V2video generationproduced by

Gen-4video generationproduced by

MAGI-1video generationproduced by

Wan 2.1 T2V 1.3Bvideo generationproduced by

Step-Video-T2Vvideo generationproduced by

Ray2video generationproduced by

Veo 2video generationproduced by Ruyi-Mini-7Bvideo generationproduced by

Mochi 1video generationproduced by

Pyramid Flowvideo generationproduced by

Video-01video generationproduced by

SVD img2vid-XT 1.1video generationproduced by

HunyuanVideovideo generationproduced by

Stable Video Diffusionvideo generationproduced by

SVD img2vidvideo generationproduced by

SVD img2vid-XTvideo generationproduced by

I2VGen-XLvideo generationproduced by Zeroscope v2video generationproduced by

ModelScope Text-to-Video (ModelScopeT2V)video generationproduced by ABot-World v0.1video generationproduced by

Wanvideo generationproduced by

Wan 2.5video generationproduced by

Wan 2.6video generationproduced by

Wan 2.7video generationproduced by

Seedancevideo generationproduced by

Seedance 1.5 Provideo generationproduced by

Lucy 2.0video generationproduced by

Veovideo generationproduced by

Veo 3.1 Fastvideo generationproduced by

Gemini Omnivideo generationproduced by

HiDream-O1-Videovideo generationproduced by IPOCvideo generationproduced by IPOWvideo generationproduced by JT-CVvideo generationproduced by JT3.5video generationproduced by

Klingvideo generationproduced by

Kling 1.6video generationproduced by

Kling 2.0video generationproduced by

LTXvideo generationproduced by

LTX-2.5 Fastvideo generationproduced by

LTX-2.5 Provideo generationproduced by

LTX-Videovideo generationproduced by

Ray 3.2video generationproduced by

Hailuo / MiniMax videovideo generationproduced by

Hailuo-02video generationproduced by

Soravideo generationproduced by

Sora 2video generationproduced by

Sora 2 Provideo generationproduced by

Soravideo generationproduced by

Pikavideo generationproduced by

Aleph 2.0video generationproduced by

Runway Genvideo generationproduced by

Gen-3video generationproduced by

Gen-4.5video generationproduced by

Viduvideo generationproduced by

Vidu Q2video generationproduced by

Vidu Q3 Turbovideo generationproduced by

Vidu Svideo generationproduced by

Grok Imagine Video 720pvideo generationproduced by

HunyuanVideovideo generationproduced by ToMoviee 2.0video generationproduced by

Amazon Nova Reelvideo generationproduced by

Leonardo Motion 2.0video generationproduced by

Ray Flash 2video generationproduced by

P-Videovideo generationproduced by

Links

R&D TopicText-to-Video Foundation Models

Models

Wan 2.1Alibaba (Qwen), video generation

Veo 3Google DeepMind, video generation

Kling 3.0Kuaishou, video generation

HunyuanVideo 1.5Tencent, video generation+100 more

Use Casesnone recorded

Who Runs Itnone recorded

Graph

Models 104

WanAlibaba (Qwen), video generationproduced by

Gen-4.5Runway, video generationproduced by

KlingKuaishou, video generationproduced by

ViduShengShu Technology, video generationproduced by

Wan 2.2Alibaba (Qwen), video generationproduced by

Hailuo / MiniMax videoMiniMax, video generationproduced by

LTX-2.5Lightricks, video generationproduced by

Vidu Q3 ProShengShu Technology, video generationproduced by

Kling 3.0Kuaishou, video generationexemplar

LTXLightricks, video generationproduced by

MiniMax H3 (Hailuo 3)MiniMax, video generationproduced by

SeedanceByteDance, video generationproduced by

+92

Seedance 2.0ByteDance, video generationproduced by

Stable Video DiffusionStability AI, video generationproduced by

Veo 3.1Google DeepMind, video generationproduced by

Gemini Omni FlashGoogle DeepMind, Google, video generationproduced by

Hailuo 2.3MiniMax, video generationproduced by

Hailuo-02MiniMax, video generationproduced by

Kling 2.0Kuaishou, video generationproduced by

Kling 2.1Kuaishou, video generationproduced by

LTX-2Lightricks, video generationproduced by

Runway GenRunway, video generationproduced by

Seedance 1.5 ProByteDance, video generationproduced by

Sora 2OpenAI, video generationproduced by

VeoGoogle DeepMind, video generationproduced by

Vidu Q2ShengShu Technology, video generationproduced by

Vidu Q4 PreviewShengShu Technology, video generationproduced by

Wan 2.1Alibaba (Qwen), video generationexemplar

Wan 2.5Alibaba (Qwen), video generationproduced by

Wan 2.6Alibaba (Qwen), video generationproduced by

Wan 2.7Alibaba (Qwen), video generationproduced by

FLUX 3 VideoBlack Forest Labs, video generationproduced by

Gemini OmniGoogle, video generationproduced by

Gemini Omni Flash 1.1Google, video generationproduced by

Gen-4Runway, video generationproduced by

Grok Imagine Video 1.5SpaceXAI (xAI), video generationproduced by

Grok Imagine Video 720pSpaceXAI (xAI), video generationproduced by

HunyuanVideoTencent, video generationproduced by

HunyuanVideo 1.5Tencent, video generationexemplar

Kling 1.6Kuaishou, video generationproduced by

LTX-VideoLightricks, video generationproduced by

MiniMax H3 MaxMiniMax, video generationproduced by

Seedance 1.0ByteDance, video generationproduced by

Seedance 2.5ByteDance, video generationproduced by

SoraOpenAI, video generationproduced by

Veo 3Google DeepMind, video generationexemplar

Video-01MiniMax, video generationproduced by

Vidu Q1ShengShu Technology, video generationproduced by

Vidu Q3 TurboShengShu Technology, video generationproduced by

Wan 2.1 T2V 1.3BAlibaba (Qwen), video generationproduced by

Wan 3.0Alibaba (Qwen), video generationproduced by

Amazon Nova ReelAmazon, video generationproduced by

Gen-3Runway, video generationproduced by

HappyHorse 1.0Alibaba (Qwen), video generationproduced by

HappyHorse 1.1Alibaba (Qwen), video generationproduced by

Kling 2.5 TurboKuaishou, video generationproduced by

Kling 3.0 OmniKuaishou, video generationproduced by

LTX-2.5 FastLightricks, video generationproduced by

LTX-2.5 ProLightricks, video generationproduced by

Lucy 2.0Decart, video generationproduced by

MAGI-1Sand AI, video generationproduced by

MAGI-2 PreviewSand AI, video generationproduced by

PikaPika (Pika Labs), video generationproduced by

Pika 2.5Pika (Pika Labs), video generationproduced by

Pyramid FlowPeking University, Kuaishou, video generationproduced by

Sora 2 ProOpenAI, video generationproduced by

SVD img2vidStability AI, video generationproduced by

SVD img2vid-XTStability AI, video generationproduced by

SVD img2vid-XT 1.1Stability AI, video generationproduced by

Veo 2Google DeepMind, video generationproduced by

Veo 3.1 FastGoogle DeepMind, video generationproduced by

Veo 3.1 LiteGoogle DeepMind, video generationproduced by

Vidu SShengShu Technology, video generationproduced by

Wan 2.2 A14BAlibaba (Qwen), video generationproduced by

Wan 2.2 TI2V-5BAlibaba (Qwen), video generationproduced by Agnes-Video-2.5Sapiens AI, video generationproduced by

Aleph 2.0Runway, video generationproduced by

Bernini-RByteDance, video generationproduced by

HiDream-O1-VideoHiDream.ai, video generationproduced by

I2VGen-XLAlibaba (Qwen), video generationproduced by

Leonardo Motion 2.0Leonardo.Ai, video generationproduced by

LongCat-VideoMeituan, video generationproduced by

Mochi 1Genmo, video generationproduced by

ModelScope Text-to-Video (ModelScopeT2V)Alibaba (Qwen), video generationproduced by

Muse VideoMeta, video generationproduced by

P-VideoPruna AI, video generationproduced by

PixVerse V6AISphere (PixVerse), video generationproduced by

Ray 3.2Luma AI, video generationproduced by

Ray Flash 2Luma AI, video generationproduced by

Ray2Luma AI, video generationproduced by

SkyReels V4Skywork AI, video generationproduced by

SkyReels-V2Skywork AI, video generationproduced by

Step-Video-T2VStepFun, video generationproduced by

Utopai XUtopai Studios, video generationproduced by ABot-World v0.1video generationproduced by IPOCvideo generationproduced by IPOWvideo generationproduced by JT-CVvideo generationproduced by JT3.5video generationproduced by Ruyi-Mini-7Bvideo generationproduced by ToMoviee 2.0video generationproduced by Zeroscope v2video generationproduced by

Papers

Imagen Video: High Definition Video Generation with Diffusion ModelsJonathan Ho et al., arXiv preprint, 2022 founding
Wan: Open and Advanced Large-Scale Video Generative ModelsTeam Wan et al., arXiv preprint, 2025 state of the art
CogVideoX: Text-to-Video Diffusion Models with An Expert TransformerZhuoyi Yang et al., ICLR2025 (per arXiv comment), 2024 state of the art
Align your Latents: High-Resolution Video Synthesis with Latent Diffusion ModelsAndreas Blattmann et al., arXiv preprint, 2023 state of the art
Video Diffusion Models: A SurveyAndrew Melnik et al., arXiv preprint, 2024 survey

Where this comes from

The papers to read first: where the direction began, its current best, and a survey when one exists.

Sources

arxiv.org/abs/2503.20314 https://arxiv.org/abs/2503.20314

To cite this page: Project AURION, AI Use-Cases Research & Innovation, "Text-to-Video Foundation Models", https://aurion.semai.id/topics/text-to-video-foundation-models/, records of October 2026.