jd-opensource/JoyAI-Echo

Proper Comfyui support with LTX2.3 is necessary. I can't test this.
I used my own comfyui ltx2.3 workflow this this model, it's not working well especially upscaler.
Thank you for your devotion!

Thanks for your interest!

The base undistilled model is planned to be released in the future. It is currently under internal company review, and we will update the repository once the release process is completed.

Thanks for your patience and support!

> Hi, thank you for your interest in JoyAI-Echo and for your feedback.
>
> JoyAI-Echo currently requires the official inference script and configuration provided in this repository. Its generation quality depends on our **DMD-distilled few-step inference** settings and the **paired audio-video memory** mechanism used in the **multi-shot long-video pipeline**. Therefore, directly using a generic LTX 2 / LTX 2.3 ComfyUI workflow or default ComfyUI parameters may not reproduce the results shown in our official examples.
>
> At the moment, we recommend using the official inference entrypoint:
>
> `python inference.py --config configs/inference.yaml`
>
> The key recommended settings are:
>
> * Resolution: 1280 × 736
> * Frames: 241
> * FPS: 25
> * dtype: bfloat16
> * **DMD denoising steps**: [1000, 994, 988, 981, 975, 909, 725, 422, 0]
> * **DMD denoising sigmas**: [1.0, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875, 0.0]
> * **Audio-video memory**: enabled
> * **Memory ma...

Hi, thank you for your interest in JoyAI-Echo and for your feedback.

JoyAI-Echo currently requires the official inference script and configuration provided in this repository. Its generation quality depends on our **DMD-distilled few-step inference** settings and the **paired audio-video memory** mechanism used in the **multi-shot long-video pipeline**. Therefore, directly using a generic LTX 2 / LTX 2.3 ComfyUI workflow or default ComfyUI parameters may not reproduce the results shown in our official examples.

At the moment, we recommend using the official inference entrypoint:

`python inference.py --config configs/inference.yaml`

The key recommended settings are:

- Resolution: 1280 × 736
- Frames: 241
- FPS: 25
- dtype: bfloat16
- **DMD denoising steps**: [1000, 994, 988, 981, 975, 909, 725, 422, 0]
- **DMD denoising sigmas**: [1.0, 0.99375, 0.9875, 0.98125, 0.975, 0.909375, 0.725, 0.421875, 0.0]
- **Audio-video memory**: enabled
- **Memory max size**: 7
- **Number of fixed memo...