JoyAI-Echo's performance on the LTX-2.5 dev transformer, with nothing baked on top. These are for people who want to apply their own distilled LoRA at their own strength - or run the model slow, as a dev model.
Not few-step on their own. At 8 steps and cfg 1 these render soft and wrong. You either add a distilled LoRA (below) and run the normal distilled schedule, or run a dev recipe. If you just want to render, take the ready-made few-step files instead - Joy-LTX 2.5 on Civitai, or on Hugging Face GGUF / comfy-native. Those are exactly these dev merges with the official distilled LoRA baked in at 0.5.
What is in the merge
JoyAI-Echo (a fine-tune of LTX-2.3) is the better actor - lip-sync, expression, a voice that stays put. LTX-2.5's transformer is shape-identical to 2.3's, so Echo's delta against 2.3-dev transplants cleanly onto 2.5-dev: 070T30 = 0.7 x the video attention / feed-forward delta, 0.3 x on the modulation tables (the default dose); 100T50 = 1.0 x / 0.5 x (livelier, hotter). Nothing retrained, no distillation applied - the file is dev + delta, full stop.
How to use them
Your own distilled LoRA (the normal way). Put a Load LoRA (model only) node between the model loader and the sampler with the official
ltx-2.5-22b-distilled-lora-450-bf16.safetensors(Lightricks/LTX-2.5 on Hugging Face,loras/), then run the distilled 8-step schedule at cfg 1. 0.4-0.6 is where the picture keeps its natural grade - measured on the same seed, 0.4 / 0.5 / 0.6 are indistinguishable by eye and all cleaner than 1.0; 1.0 is fully distilled and hotter. 0.5 is what the ready-made files use.Any other distill / style LoRA the same way. That is the point of this page: the dial is yours.
As a dev model. 25-30 steps, the LTXV scheduler, cfg about 3. Slower and softer, and the Echo delta reads as extra guidance at cfg > 1 - expect it hotter than stock dev; keep cfg modest.
LoRA on the int8 file works in stock ComfyUI - it is how Lightricks ship LTX-2.5 dev + distilled LoRA. The ComfyUI-JoyLTX25 canvases (Civitai: Joy-LTX 2.5 workflows) take these files too - drop the LoRA node after the loader.
Files (one version per dose)
LTX25dev-echoVid-<dose>-DiT-comfy-int8.safetensors- 21.5 GB. Stock Load Diffusion Model; fastest on RTX 50, works on RTX 30/40 (ComfyUI 0.32+).LTX25dev-echoVid-<dose>-DiT-comfy-fp8.safetensors- 21.5 GB. comfy fp8_e4m3fn, scaled.LTX25dev-echoVid-<dose>-DiT-bf16.safetensors- 42.0 GB. The master, for merging or your own quantisation (if the upload here is refused for size it is on Hugging Face: joyai-echo-ltx25-echoVid-dev).
No GGUF on this page on purpose: a LoRA on a GGUF weight is dequantised per step and slow; quantise after you have baked your LoRA, or use the ready-made GGUF page.
You also need (Lightricks/LTX-2.5 on Hugging Face)
ltx-2.5-video-vae-bf16.safetensorsandltx-2.5-audio-vae-bf16.safetensors→models/vae/text encoder
gemma4-12b-with-proj-ltx-2.5-comfy-int8-convrot.safetensors→models/text_encoders/the distilled LoRA above →
models/loras/; the x2 latent upscaler if you run two passes.
Why this exists
The first Joy-LTX 2.5 build applied Echo's delta to the distilled 2.5 transformer and came out over-cooked - +16% saturation, hard contrast, blown highlights on any bright scene. Diagnosing it showed the delta behaves like extra guidance on a few-step model, and that a dev-base merge with the distill LoRA held back to 0.5 reads clean. That number suits my eye; these files let you pick yours.
Credits
JoyAI-Echo by JD (jdopensource/JoyAI-Echo). LTX-2.5 and the distilled LoRA by Lightricks. Merge and files by me. Licensed under the LTX-2.x Community License (inherited from both parents). Questions: comment here or open a discussion on the Hugging Face page - I answer.
Description
070T30 - the default dose - the LTX-2.5 dev transformer with JoyAI-Echo's video attention / feed-forward delta transplanted in, dosed at 0.7 on the video attention / feed-forward tensors and 0.3 on the modulation tables. Nothing is baked on top: these are not few-step files. You add the distillation, at the strength you want.
Use it one of two ways
Few-step (what most people want). Load
ltx-2.5-22b-distilled-lora-450-bf16.safetensors(Lightricks/LTX-2.5 on Hugging Face,loras/) with a Load LoRA (model only) node between the model and the sampler, then run the normal distilled 8-step schedule at cfg 1. 0.4-0.6 keeps the picture's natural grade (0.5 is what the ready-made Joy-LTX 2.5 files use); 1.0 is fully distilled and hotter. Any other distill LoRA works the same way.As a dev model. 25-30 steps, the LTXV scheduler, cfg about 3. Slower and softer - and the Echo delta reads as extra guidance above cfg 1, so keep cfg modest or it warms up fast.
The three files
LTX25dev-echoVid-070T30-DiT-comfy-int8.safetensors- 21.5 GB. Stock Load Diffusion Model; fastest on RTX 50, fine on RTX 30/40 (ComfyUI 0.32+). Start here.LTX25dev-echoVid-070T30-DiT-comfy-fp8.safetensors- 21.5 GB. comfy fp8_e4m3fn, scaled - the alternative if int8 misbehaves on your card.LTX25dev-echoVid-070T30-DiT-bf16.safetensors- 42.0 GB. The master: for merging, for your own quantisation, or a 48 GB+ card. LoRA on bf16 is the cleanest of the three.
This is the dose the released few-step files are built from and the one to start with: cleaner skin on quiet scenes, close-ups and interviews, and the most natural grade of the two.
Also needed from Lightricks/LTX-2.5: both VAEs, the Gemma-4 text encoder, and the distilled LoRA above. Workflows: ComfyUI-JoyLTX25 (drop a LoRA node after the loader - the PLUS canvases already have a four-slot LoRA stack there). Ready-made few-step files: Joy-LTX 2.5 on Civitai, or GGUF / comfy-native on Hugging
FAQ
Comments (3)
The model is heavy for 12 gb vram on 4070TI. takes lot of time to render also its not much accurate for hands and eyes.
You would do better with one of these:
Joy-LTX 2.5 Distilled GGUF / INT8 / NVFP4 / W4A8 / MIX4x8 - 16GB | LTX Video Checkpoint | Civitai
@joeygambino GGUF is too bad quality, better just use Minimax H3, it creates 15 seconds on 4070ti without troubles.