Realtime models
Lightricks LTX Video
Open latent-diffusion video model that generates clips faster than real time — the engine several chat-driven live loops run on.
github.com · listing ≠ endorsement · re-checked 2026-09-03
- Operator
- Lightricks
- Params
- ~1.9B open weights
- Model
- Latent diffusion · 1:192 Video-VAE
- Output
- Faster-than-playback clips (paper-reported)
- Access
- Open source (code + weights)
What is Lightricks LTX Video?
LTX Video is the open-weight latent-diffusion model whose paper is literally titled 'Realtime Video Latent Diffusion' — it renders fixed-length clips faster than playback (per the paper, 5s of 24fps 768×512 in ~2s on an H100). Later LTX-2/LTX-2.5 releases added audio-video sync and faster distilled variants; the line is also the generator underneath chat-driven live loops such as Infinite TV.
What Lightricks LTX Video does
- Open-weight latent-diffusion video model built around a 1:192 pixel-to-token Video-VAE
- Generates fixed-length clips faster than playback — the paper reports 5s of 24fps 768×512 in ~2s on an H100
- Later LTX-2 / LTX-2.5 releases added audio-video sync and faster distilled variants
- Acts as the generator underneath open chat-driven live loops such as Infinite TV
- Code and pretrained weights are publicly available
Who is Lightricks LTX Video for?
- Running open faster-than-playback generation on your own hardware
- Building the video layer of a chat-to-clip live stream
- Comparing an open diffusion engine with open autoregressive realtime models
Pros and cons of Lightricks LTX Video
Pros
- Genuinely open source — code and weights both available
- Proven in the wild as the engine of community live-streaming projects
- Fast enough that live loops are practical rather than theoretical
Cons
- Generates fixed-length clips, so “realtime” means fast clips, not an interactive streaming model
- Faster-than-realtime figures are the authors’ own benchmarks
- Top-tier speed assumptions (H100-class) may not hold on consumer GPUs
What people are saying
Verbatim quotes from public coverage — collected 2026-09-03.
“5 seconds of 24 fps video at 768×512 resolution in just 2 seconds on an Nvidia H100 GPU.”
Frequently asked questions
Is LTX Video actually realtime?
It generates fixed-length clips faster than playback — per the paper, 5s of 24fps video in ~2s on an H100 — rather than streaming interactively frame-by-frame.
Is LTX Video open source?
Yes — code and pretrained weights are publicly available from Lightricks.
What is LTX Video used for in live streaming?
It is the generation engine in community live loops like Infinite TV, which feeds it chat-derived prompts.
MiniMax H3
Open-weight 33B video model (Hailuo 3.0) — 15s clips with native stereo audio; the substrate under this wave.huggingface.co
fal H3 Max Live
fal’s realtime-tuned H3 Max — five seconds of 768p in under three, faster than playback.fal.ai
Krea Realtime 14B
Open 14B autoregressive model for real-time, long-form streaming video — distilled from Wan 2.1, ~1s to first frame.krea.ai
Decart MirageLSD
“Live-stream diffusion” — restyle a live feed (camera, game, call, screen) frame-by-frame from a text prompt at near-zero latency.decart.ai