Ryan Li

[AI奶爸 @zstmfhy]

一直以来，我都知道自己在codex方面的能力很弱，正好需要一个教程，苍何大佬就开源了出来，有需要的兄弟也可以点个star，随时查阅

[apolinario (poli) @multimodalart]

Whoa, this is such a cool technique!

It adds reference structure/style guidance without an inference slowdown to any flow matching model

Just built a demo for it on @huggingface Spaces

[Wildminder @wildmindai]

ComfyUI wrapper for LiTo.
Single RGBA image to 3D Gaussians in ComfyUI

https://t.co/RqR8v92DA7

[Wildminder @wildmindai]

LTX2.3 OmniNFT RL-LoRA
generates high-quality video/audio + visuals and sound are perfectly synchronized, no laggy or mismatched audio

- realistic Lip-Sync
- action-matched sound
- reduces synchronization errors by 52%
really nice output

https://t.co/MNut4JPPtE

[⚡AI Search⚡ @aisearchio]

This is the future of gaming

DreamX-World - new interactive world generator based on Wan

https://t.co/XnXjln2zcR

[Feiteng @FeitengLi]

其实这些问题都能很好的解决了

1. 扔掉 whisper，换 ASR 模型，Qwen3-ASR 就很不错幻觉很少、也有一些别的ASR选择，whisper 幻觉多也要求 30s片段，Qwen3-ASR 塞更长的音频识别越准确，最大支持 20 分钟；
2. 文字时间轴 也扔掉 whisper 不是很准, 虽然 Qwen/Qwen3-ForcedAligner-0.6B 也能用，但实际测试超过 180s 就时间轴就混乱不堪，可以用 @LattifAI_HQ https://t.co/ukwzM7NMeS 4 小时都轻松准确，https://t.co/zNDkOQZrui 可以到这里看看 卡拉 OK 字幕都十分准确，也有 skill https://t.co/bR8Nw1ahze speaker diarization 和 naming 也解决的很好了
3. VAD 切片我推荐自己的项目 https://t.co/l5yvfDJUOs 准确性 Top

[R@aiaicreate @aiaicreate]

ComfyUI向けタイムラインエディタ「LTX Director」公開。
LTX2.3での動画生成において、細かなプロンプト制御や音声追加、I2V等を単一ノードで統合。
直感的な操作性が注目を集め、複数フレーム管理を大幅に簡略化するツールとの評価。

#ComfyUI #LTXVideo

URLはリプ⬇️

[Wildminder @wildmindai]

NVIDIA just dropped another cool world model - SANA-WM.
You give it one image and a camera path, and it spits out a full minute of 720p on a single GPU.

> handles a full 60s
> 1-minute clip in just 34 seconds (on a 5090).
> follows precise 6-DoF paths, movement feels intentional and grounded.
> Based on SANA 2.6B

World generator with cinematic fly-throughs for any fantasy setting

https://t.co/N6KxXv0LL7

[Wildminder @wildmindai]

Wan2.2 again.
SwiftI2V: Efficient 2K I2V video gen with 21GB VRAM.
- uses 200x less GPU-time than CineScale
- exact image fidelity
- decoupled processing
no models yet.
https://t.co/UmfRrwq3IY

[歸藏(guizang.ai) @op7418]

https://t.co/ip9kyn2Trk

[Ryan Li @RyanLi1992]

@0x0funky 这熟悉的hyperframes视频

[Ryan Li @RyanLi1992]

@oxniubi plus会补货么，买的快到期了。。。

[Zain Shah @zan2434]

Imagine every pixel on your screen, streamed live directly from a model. No HTML, no layout engine, no code. Just exactly what you want to see.

@eddiejiao_obj, @drewocarr and I built a prototype to see how this could actually work, and set out to make it real. We're calling it Flipbook. (1/5)

[Ryan Li @RyanLi1992]

@InSpatio_AI anyone tried this on 5090 ？

[Ryan Li @RyanLi1992]

4/ Feels like browser-operator AI + local AI tools is an underrated combo.
Curious if anyone else is experimenting with this.

[Ryan Li @RyanLi1992]

1/
You might not need Qwen-VL anymore for image→video prompts.
I found a free trick using Gemini in Chrome that can auto-caption images and run my I2V workflows.

Works with tools like Wan2GP and ComfyUI.
And it worked better than I expected 👇

[Ryan Li @RyanLi1992]

3/ Result: Gemini acts like a free vision model + workflow operator.
Great workaround if you want Qwen-VL style captioning but don’t have enough VRAM.
Limitation: this is UI automation, so it won’t work for API pipelines — but for local workflows it’s surprisingly powerful.

[Wildminder @wildmindai]

ok, Video AI is finally moving from aesthetic vibes to world understanding.
VBVR- a foundational scale-up for video reasoning;
- shift focus from visual quality to spatiotemporal intelligence, reason over motion, interaction, causality.
https://t.co/53C32C2ExU

[Ryan Li @RyanLi1992]

This is what i need

[Wildminder @wildmindai]

ComfyUI-Launcher.
An Electron app for managing multiple ComfyUI installations.
https://t.co/mAsUYAKDdP