JoyAI-Echo x LTX-2.5 (echoVid) - GGUF
LTX-2.5's engine with JoyAI-Echo's performance. LTX-2.5 renders picture and sound in
one pass, at any length, in one generation. JoyAI-Echo (a fine-tune of LTX-2.3) has the
better actor: natural lip-sync, expressive faces, a voice that stays put. The two
transformers are shape-identical, so JoyAI-Echo's video attention/feed-forward delta was
transplanted onto the official LTX-2.5 dev transformer, and the official LTX-2.5 distilled LoRA (ltx-2.5-22b-distilled-lora-450) is baked in at 0.5 - so these are few-step files with the same speed, VRAM and nodes as LTX-2.5 distilled. Nothing was retrained. (v2: the first build put the delta on the distilled transformer and came out over-saturated with hard contrast; those files are gone. The plain dev merges, for people who want to apply their own distill LoRA at their own strength, are here: https://huggingface.co/joeygambino/joyai-echo-ltx25-echoVid-dev.)
What you get over stock LTX-2.5 distilled is the acting JoyAI-Echo was trained for - lip-sync, expression, a voice that stays put - at the same speed, VRAM and nodes.
Workflow + nodes: https://github.com/jlucasmcrell/ComfyUI-JoyLTX25 (the Joy-LTX 2.5 canvases: one-prompt take with a VRAM planner, and multishot with AV-extend joins; the release zip bundles the writer). comfy-native (int8 / w4a8 / w4a4 / nvfp4 / mixed) files: https://huggingface.co/joeygambino/joyai-echo-ltx25-echoVid-comfy-native All models: https://huggingface.co/joeygambino Try it live: https://huggingface.co/spaces/joeygambino/joy-ltx-25 (one take, ZeroGPU) Civitai: Joy-LTX 2.5 (models being uploaded now).
What it looks like
Rendered with the files on this page (070T30, distilled LoRA baked at 0.5), the ComfyUI-JoyLTX25 canvases, 8 steps at cfg 1. Sound is generated with the picture, in the same pass - turn it on.
Three shots joined into one take
Multishot, 3 x 8 s at 1280x736, AV-extend joins - the speech and the room carry across both joins with no reference photo attached.
Beach, hard sun
10 s, single generation, picture and sound together.
Wet neon street
10 s, single generation. Reflections and rain with a voice over them.
Snow, flat overcast
10 s, single generation. The grade holds in high key - the failure mode of the first build.
Two doses
| dose | what it is | pick it when |
|---|---|---|
| 070T30 (default) | 0.7 x Echo delta on video attention/FF, 0.3 x on the modulation tables, distill LoRA 0.5 | the default - cleaner skin, natural grade |
| 100T50 (strong) | 1.0 x / 0.5 x, distill LoRA 0.5 | loud, comic, animated performances - the livelier read, a touch hotter on contrast |
Both were reviewed blind on 20+ paired renders: scores tie; 070T30 reads a touch less rubbery on still faces, 100T50 lands laughter and big expressions better. Start with 070T30.
Which file (24 GB and under, GGUF is the fast family on RTX 30/40)
Sizes are decimal GB. Timings: 960x544, 8 s, two-pass x2 to 1920x1088, RTX 3090.
| file | GB | fits | 3090 per 8 s clip |
|---|---|---|---|
LTX25dist-echoVid-<dose>-v2-DiT-Q3_K_M.gguf | 10.6 | 12 GB (tight) / 16 GB | - |
LTX25dist-echoVid-<dose>-v2-DiT-Q4_K_S.gguf | 12.9 | 16 GB | 491 s |
LTX25dist-echoVid-<dose>-v2-DiT-Q4_K_M.gguf | 14.2 | 16 GB (tight) | 484 s |
LTX25dist-echoVid-<dose>-v2-DiT-Q5_K_M.gguf | 15.9 | 24 GB | 466 s (24 GB default) |
LTX25dist-echoVid-<dose>-v2-DiT-Q6_K.gguf | 17.7 | 24 GB | 399 s |
LTX25dist-echoVid-<dose>-v2-DiT-Q8_0.gguf | 22.7 | 32 GB | - |
On an RTX 5090 the comfy-native files are 1.5-2x faster than GGUF (see the other repo).
Install (ComfyUI)
- Install ComfyUI-GGUF (city96) from the Manager if you have not.
- Put the
.ggufinmodels/diffusion_models/. - From Lightricks/LTX-2.5:
vae/ltx-2.5-video-vae-bf16.safetensorsandvae/ltx-2.5-audio-vae-bf16.safetensors->models/vae/;latent_upscale_models/ltx-2.5-latent-spatial-upscaler-x2-bf16-1.0.safetensors->models/latent_upscale_models/; text encodertext_encoders/gemma4-12b-with-proj-ltx-2.5-comfy-int8-convrot.safetensors->models/text_encoders/(16 GB cards: the 10.6 GBgemma4-12b-ltx25-comfy-w4a8.safetensorsfrom LTX-2.5-Quantized). - Load the workflow from the node pack above (or any LTX-2.5 workflow: swap the loader for
UnetLoaderGGUF and pick this file). Distilled schedule: 8 steps pass 1, 3 steps pass 2,
euler_ancestral, CFG 1.
Credits
JoyAI-Echo by JD (jdopensource/JoyAI-Echo); LTX-2.5 by Lightricks. Merge, quantisation and workflows by joeygambino. Licensed under the LTX-2.x Community License (inherited from both parents).