Upload folder using huggingface_hub

.gitattributes

@@ -42,3 +42,14 @@ examples/example_6_reference.mp4 filter=lfs diff=lfs merge=lfs -text
 examples/example_7_reference.mp4 filter=lfs diff=lfs merge=lfs -text
 examples/example_8_reference.mp4 filter=lfs diff=lfs merge=lfs -text
 examples/example_9_multi_task.mp4 filter=lfs diff=lfs merge=lfs -text
+examples/refs/084b67fac2342669a7802a282dfb397d86279f8854a683f6ea183544561e1a7b.png filter=lfs diff=lfs merge=lfs -text
+examples/refs/10cac154a98bf7c4b6320787c59563d63dcbb74a8f07f5e8d4229c8a2bb953c3.png filter=lfs diff=lfs merge=lfs -text
+examples/refs/35e764bd898b6556d59bf91c14f8634db64432034b8a35e099ad1645e6e75ab6.png filter=lfs diff=lfs merge=lfs -text
+examples/refs/3e57d239e1a327c569e518b748d84761c7301bf79fb82ef1d266040bc122bb62.png filter=lfs diff=lfs merge=lfs -text
+examples/refs/3e5806092fab2571046bca535026065b852e8708463b224dac363f8992a0eba2.png filter=lfs diff=lfs merge=lfs -text
+examples/refs/3ea5f738336d6afa6aca8b47d0d0288ccf53d7c334f623896b91d4c9bc47d07f.png filter=lfs diff=lfs merge=lfs -text
+examples/refs/4a1c6d4c36e7b442890ca061cffc9e2c1bb2aca8ec4ae1128750752cf8ddabff.png filter=lfs diff=lfs merge=lfs -text
+examples/refs/64ada2c934c988feea57f0597366684332e9ea4f25c8b2a96b779ab0d17664e4.png filter=lfs diff=lfs merge=lfs -text
+examples/refs/7dce66ebc96daa0dc0ee9eea16606406b745d83ff0bc6bce3eff32865603dd13.png filter=lfs diff=lfs merge=lfs -text
+examples/refs/7e43a6c8b433fed0f1382feb50a60482cdc159448b3191edfbad51301e588e31.png filter=lfs diff=lfs merge=lfs -text
+examples/refs/80eb8745da114b62ae825a681332159792402dca88fb3e5a57ad7ed340572ed8.png filter=lfs diff=lfs merge=lfs -text

README.md

@@ -6,7 +6,6 @@ tags:
   - lora
   - video
   - video-editing
-  - ltxv
   - ltx-2.3
 ---
 
@@ -134,18 +133,25 @@ conditioning — the model is driven entirely by the text prompt and the
 guide video. Trained on a balanced mix of **Add, Remove, Replace, Style**
 edits.
 
-### What it's good for
+### What it's different about it (vs v0.1)
 
-- Multi-task edits without dealing with reference images.
-- **Remove** and **Style** specifically — those aren't on v0.1's menu.
-- Cases where preparing an edited first frame (the motion-transfer route)
-  isn't an option.
+The task vocabulary overlaps heavily with v0.1 — both can do Add, Remove,
+Replace, Change, Convert. What changes here:
+
+- **Two-stage training continuation**: the first stage gave the model its
+  edit vocabulary; the second stage refined it on a larger, more balanced
+  video pair set covering Add / Remove / Replace / Style.
+- **Rank 256** (vs v0.1's effective rank from the merge), giving more
+  capacity for the broader task mix.
+- Trained directly on video pairs, so the temporal behaviour on these
+  tasks tends to be steadier than on a model whose first stage was on
+  still images.
 
 ### How to use it
 
-This LoRA was fine-tuned on top of v0.1's lineage, so it expects v0.1 to be
-loaded with it. **Stack v0.1 + this LoRA together** at inference. If
-results look over-saturated, drop v0.1's strength to ~0.7 first.
+**Standalone** — load it as a regular LoRA on vanilla LTX-2.3 through any
+ComfyUI LoRA loader. The file already carries everything it needs; no
+stacking with v0.1, no companion module.
 
 ### Limitations
 
@@ -159,9 +165,9 @@ Same imperative shape as v0.1, but the training set split into four very
 distinct caption styles. Match the one that fits the edit you want — the
 distribution is narrow and the model expects the right shape.
 
-The dataset has **~6000 Add, 6000 Remove, 6000 Replace, ~3840 Style**
-samples (~22k total). Captions below are real examples drawn from those
-buckets.
+The training set is roughly balanced across **Add, Remove, Replace and
+Style** buckets, with Style being the smallest of the four. Captions
+below are real examples drawn from those buckets.
 
 #### Add — 15 to 30+ words, describe what to add and where
 
@@ -219,11 +225,10 @@ These are honest limits of the training distribution — don't expect them
 to work just because the model is multitask:
 
 - **No compositional prompts.** *"Add X and remove Y"*, *"Replace A with B
-  and add C"*, etc. are **not** in the training set. Out of ~22k captions,
-  only 13 combine two action verbs at all, and almost all of those are
-  the "Remove X and replace with Y" idiom (which is really a single
-  Replace). Pure multi-action edits will fall apart — split them into
-  separate runs.
+  and add C"*, etc. are **not** in the training set. Captions combining
+  two action verbs are essentially absent (the only ones present are the
+  "Remove X and replace with Y" idiom, which is really a single Replace).
+  Pure multi-action edits will fall apart — split them into separate runs.
 - **No "change background" as a task.** Background is only used as a
   *positional reference* ("in the background", "on the wall in the
   background"). To swap the entire backdrop, phrase it as a **Replace**
@@ -240,14 +245,13 @@ to work just because the model is multitask:
 | If you want… | Use |
 |---|---|
 | Motion transfer (edit first frame externally, model copies motion) | **v0.1 motion transfer** |
-| Multi-task edits (add / remove / replace / style) driven only by prompt | **no-ref multitask r256** (stacked with v0.1) |
+| Multi-task edits (add / remove / replace / style) driven only by prompt | **no-ref multitask r256** (standalone) |
 | Strong identity transfer from a reference image (Add / Replace) | **Ref V2V** |
 
 ### How to wire the LoopingSampler
 
-Two standard LoRAs stacked, no sidecar. Chain through two ComfyUI
-LoraLoaders before the LoopingSampler: load **v0.1** first, then the
-**no-ref multitask** LoRA on top. On the sampler:
+A single **standard LoRA**, no sidecar, no stacking. Load through one
+ComfyUI LoraLoader before the LoopingSampler. On the sampler:
 
 - `editanything_module` → **leave disconnected**.
 - `ref_image` → **leave disconnected**. This LoRA has no reference-image
@@ -463,6 +467,44 @@ Tips for Ref V2V:
 
 ---
 
+## Inference tips (applies to all models)
+
+**CFG matters a lot here.** The default workflow runs with the LTX-2.3
+**distilled / acceleration LoRAs** for fast 4–8 step sampling, which
+locks **CFG = 1.0**. That's fine for casual runs, but at CFG 1 the model
+follows the prompt loosely — you get the reference image to "show up"
+but the edit instruction itself is only weakly enforced.
+
+**For harder prompts, raise CFG above 1.0.** This means dropping (or
+weakening) the distilled / acceleration LoRAs and going back to a normal
+sampler with more steps — significantly slower, but the model follows
+the prompt much more closely. Trade-off:
+
+| Setup | CFG | Steps | Speed | Prompt following |
+|---|---|---|---|---|
+| Distilled / accel default | **1.0** | 4–8 | fast | loose |
+| Non-distilled, mild | **3–4** | 20–30 | medium | good baseline |
+| Non-distilled, aggressive | **5–7** | 25–40 | slow | strong, may saturate |
+
+Other knobs:
+
+- If the model is **ignoring the prompt** (edit isn't being applied, the
+  reference is barely showing up, the style transfer is faint), raising
+  CFG is the single most common fix. Go up to 6–8 if needed.
+- If results look **over-saturated, plasticky, or motion is freezing**,
+  CFG is too high — pull back toward 3–4 or re-enable the distilled LoRA
+  for CFG 1 if you don't actually need stronger prompt adherence.
+- Ref V2V in particular benefits from being more aggressive with CFG when
+  the reference identity isn't transferring cleanly.
+- Combine CFG tuning with the LoRA-specific knobs from each section
+  (`adaln_scale`, `ref_context_scale`, `ref_token_scale` for Ref V2V;
+  prompt rewriting for v0.1 / no-ref).
+
+Treat CFG as a real knob, not a constant — and be ready to give up some
+speed when you actually need the edit to land.
+
+---
+
 ## ComfyUI nodes
 
 All recommended inference paths run through the **BFSnodes** custom node
@@ -471,6 +513,7 @@ stabilise they may move elsewhere.
 
 Specific nodes used by these LoRAs:
 
+- `LTXVEditAnythingApply` — load the LoRA + extras and patch the model.
 - `🅛🅣🅧 LTXV Edit Anything Looping Sampler` — sampler that injects role /
   AdaLN / visual cross-attention and handles long videos in chunks.
 - `LTXVEditAnythingModuleLoader` — load the `*.module.safetensors` sidecar.