eisneim

Gen AI | Web developer | videographer 视频大拍档特里

[eisneim @eisneim]

I'm working on a Flux dev based model that can relight a photo conditioned on time (eg. 6AM, 7AM ) without changing the background unlike ic-light and LBM model

[Wildminder @wildmindai]

LTX-2.3 OmniCine V1 LoRA
- Anatomy fix
- Director controls:
- Better lip-sync and facial nuance + it finally stops burnt-in subtitles from ruining the video.
- Objects and characters don't warp when things get fast or chaotic.
- Handles 2D Anime, 3D CGI, photorealism.

https://t.co/i2eNhXLck8

[Wildminder @wildmindai]

LTX2.3 Seamless Transitions IC-LoRA
perfect loops every time

https://t.co/fZkruqx1lV

[Gorden Sun @Gorden_Sun]

Warp-as-History：仅用一条视频就能实现交互式视频生成
用单条带标注的视频做轻量LoRA微调后，即可让通用视频模型实现跟随视角生成视频。项目里用的这条视频是来自DAVIS数据集里的car-roundabout.mp4。
原理是：把相机轨迹产生的变形(warp)伪装成视频模型原生的"历史帧"输入，无需额外的相机编码器或控制分支，就能让预训练视频生成模型跟随指定视角运动。
Github：https://t.co/a2z3jwyS7H

[Gong Junmin @junmingong]

Khala 1.0 just dropped — a music generation model from the Central Conservatory of Music in Beijing. Paper, code, weights, and demo all open-sourced.

I gave a talk there recently on ACE-Step and got an early look at Khala. Excited to see it officially out. Open-source music gen is thriving.

💻 https://t.co/iYQt9e1mMy
📝 https://t.co/fqwqtvHfP1
🎧 https://t.co/XAxqLEYGft

[Mickmumpitz @mickmumpitz]

I've been working on a bigger AI VFX pipeline and needed audio-driven vid2vid lip sync for @LTXStudio LTX 2.3. Couldn't find a workflow for it, so I built this one in @ComfyUI.

More examples, free guide and free workflows below! 👇

[Kai He @Kai__He]

We open-sourced the code and model for UniRelight! 🎉

Given an input video and a target lighting configuration, our method jointly predicts a relit video and its corresponding albedo.

Code: https://t.co/4zF94saWvo
Model: https://t.co/d8i66UyvhU

[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

[eisneim @eisneim]

https://t.co/VVVriWJdT0 I converted SAM2.1 to mlx so you can run video segmentation on Mac using apple Silicon GPU, no cuda gpu needed

[Mickmumpitz @mickmumpitz]

Another test with the LTX 2.3 vid2vid lip sync workflow. I've been finding the inpainting mode works more reliably overall, so I'd actually recommend turning it on even for close-ups.

[Brie Wensleydale🧀🐭 @SlipperyGem]

Yet another amazing-lookingIC lora for LTX 2.3 lands on the scene.

Its v2v and text prompted. Does editing, removal, replacement and restyle.

Personally, I would REALLY like to know if it can handle a first frame as a reference. I'm guessing now though.
https://t.co/8Ymjmd0KQl

[Purz.ai @PurzBeats]

LTX 2.3 IC LoRA - EditAnything by Alisson Pereira

[A.Robot @100PercentRobot]

Just discovered frame injection in LTX-2.3, so of course I did something weird

[⚡AI Search⚡ @aisearchio]

Another open source image generator & editor

LLaDA2.0-Uni

https://t.co/i8ohSPHx1z

[eisneim @eisneim]

https://t.co/9vrV2Kkedu
i created a new Repo for faster and better image to video generation using LTX 2.3 with triple stage sampling

[eisneim @eisneim]

https://t.co/T2HPMrbtpy
new video model: 15B-parameter, 40-layer Transformer that jointly processes text, video, and audio via self-attention only. No cross-attention, no multi-stream complexity. Achieves 80.0% win rate vs Ovi 1.1 and 60.9% vs LTX 2.3

[Tongyi Lab @Ali_TongyiLab]

1/2 Qwen3.5 is here. The next frontier of Native Multimodal Agents is open. 🚀
We are thrilled to release Qwen3.5-397B-A17B, our flagship open-weight vision-language model. Built for the future of coding, reasoning, and seamless multimodal interaction.
Key Highlights:
Inference Efficiency: A massive 397B total parameters, but only 17B active—delivering flagship power at a fraction of the cost.
Hybrid Architecture: Innovative Gated Delta Networks (Linear Attention) + Sparse MoE for extreme speed.
True Multimodality: Exceptional performance across GUI interaction, video comprehension, and agentic workflows.
Global Scale: Qwen3.5 now supports over 200 languages.
Empowering developers and enterprises to build smarter, faster, and more versatile AI agents.

[Ivan Fioravanti ᯅ @ivanfioravanti]

OpenCode + MLX + Qwen3.5-397B-A17B-4bit.
Video is 8x, but the goal is showing that It works!

This is something unimaginable just few months ago.
MLX Team is pushing like crazy and M5 Ultra will do the rest 🚀

[Wildminder @wildmindai]

Capybara? 14B model for T2V, T2I, TV2V, TI2I.
- based on HunyuanVideo1.5;
- byt5-small, Glyph-SDXL-v2, SigLIP;
- 480p-1080p; 16.7GB model, 5GB VAE..
mostly for video editing.
https://t.co/N34iJ4gC0K

[Gorden Sun @Gorden_Sun]

BitDance：字节大年初一开源的AI绘画模型
最大的亮点是速度快，使用高压缩视觉分词器，将图像映射为紧凑的二值Token序列，并且每一步扩散过程并行预测64个Token。所以即使模型大小有14B，生成图片的速度也非常快。
模型：https://t.co/6FtVbAU4uk
Github：https://t.co/ux7k7xVvbA

[Wildminder @wildmindai]

Self-Refining Video Sampling: inference-time method using a video generator as its own refiner to correct physics and motion.
no retraining needed; scores >70% human preference; is validated on Wan2.2 & Cosmos.
https://t.co/NGdxcTUNeX