Alissonerdx/LTX-Best-Face-ID · Hugging Face

Source: https://huggingface.co/Alissonerdx/LTX-Best-Face-ID

https://huggingface.co/Alissonerdx/LTX-Best-Face-ID#ltx-best-face-id–ltx-2-identity-lora-reference-to-video–ipt2vLTX-Best-Face-ID — LTX-2 Identity LoRA (Reference-to-Video / IPT2V)

An identity-preservingreference-to-videoLoRA forLTX-2 (22B). Give it a reference photo of a person + a text prompt, and it generates a video that keeps that person’s identity.

Built withoverlap reference conditioning + TASS-RoPE (source-phase RoPE)and a differentiableArcFace identity loss. Runs in ComfyUI via the companionBFS Nodes.

Status:two reference modes are now available from the same recipe: the originalclose-up facereference, and a newercharacter-sheetreference (face + full-body views) that also carries clothing/appearance consistency — seeCharacter-Sheet Referencebelow. Other experimental variants (native Gemma-vision conditioning, timestep-split texture injection) may be released later if they prove out.

https://huggingface.co/Alissonerdx/LTX-Best-Face-ID#%F0%9F%8E%AC-examples🎬 Examples

https://huggingface.co/Alissonerdx/LTX-Best-Face-ID#with-2-pass-upscaledWith 2 Pass (Upscaled)

https://huggingface.co/Alissonerdx/LTX-Best-Face-ID#what-it-doesWhat it does

• **Reference-to-video (ref_t2v):**one reference image → video of that identity performing the prompt’s action.
• Identity is injected by placing thereference latentin the target’s frame-0 RoPE grid (overlap) and tagging it with a distinctsource phaseso the model knows it is areference, not the first frame to generate.
• An auxiliaryArcFace face-similarity losson the decoded prediction sharpens the identity.

https://huggingface.co/Alissonerdx/LTX-Best-Face-ID#how-it-works-techniqueHow it works (technique)

https://huggingface.co/Alissonerdx/LTX-Best-Face-ID#overlap-reference–tass-rope-source-phaseOverlap reference + TASS-RoPE (source-phase)

The reference latent is concatenated to the video sequence sharing the frame-0 grid (classic IC-LoRA “overlap”). To stop the reference from leaking into / being confused with the generated first frame, each source gets a distinctmultiplicative RoPE phase:

phase[d] = source_id · phase_scale · θ^(−d/L)      (θ = 10000)
target tokens: source_id = 0  (phase 0, exact no-op)
reference:     source_id = 2  (distinct rotary "tag")

This “source tag” lets the model separatewho is whoin the sequence and strongly improves identity transfer. Because the tag is positional, the same mechanism generalizes tomultiple references(source_id = 2, 3, 4, …) for multi-subject conditioning.

TASS-RoPE(Temporal-Adjacent Spatial-Shifted RoPE) is fromST-DRC, Chen et al.,Spatial-Temporal Decoupled Reference Conditioning for Identity-Preserving Text-to-Video Generation—arXiv:2606.02441. This LoRA uses an overlap + per-source RoPE-phase variant of that idea.

https://huggingface.co/Alissonerdx/LTX-Best-Face-ID#arcface-identity-lossArcFace identity loss

During training the flow-matching prediction is decoded to pixels, the face is aligned and passed through ArcFace (buffalo_lw600k\_r50), and a cosine identity loss pulls the generated face toward the reference embedding — the same recognition space used to evaluate identity.

https://huggingface.co/Alissonerdx/LTX-Best-Face-ID#character-sheet-reference-newCharacter-Sheet Reference (NEW)

The base model above (and the guidance further down) was trained onclose-up face crops— great for identity, but the model never saw the person’s actual clothing or body, so it has toinventan outfit rather than reproduce a specific one.

We’ve since trained acontinuation checkpointon acharacter-sheet referenceinstead: a single image with4 panels side by side — a face close-up, then full-body front / side / back viewsof the same person, all on a white background. It uses the exact same mechanism (source\_id=2,layout=overlap) — it’s just adifferent, richer reference image, so the model learns to reproduce the reference’s clothing and build consistently, not only the face.

This checkpoint continues training from the close-up FaceID LoRA above rather than replacing it— in our tests the combination (close-up identity foundation + character-sheet continuation) gave noticeably better and more consistent results than either recipe trained alone. If you only have a plain headshot, the original close-up model still works well; the character-sheet checkpoint is for when you want the generated video to also match a specific outfit/build.

File:Best\_FaceID\_CharacterSheet\_v1\.0\_LoRA\.safetensors— use it exactly like the base LoRA above (same node, samesource\_id=2/layout=overlap), just swap the reference image for a 4-panel character sheet instead of a headshot, and setref\_resize\_mode = native\_resolution(see below). This checkpoint hasno ArcFace projector— its identity/appearance comes entirely from the overlap reference latent, so leave that dropdown onNone.

💡Identity coming out weak?Load the baseBest\_FaceID\_v1\.0\_LoRA\.safetensorsalongside this one at a low strength (~0.2 or more) — mixing in some of the base close-up LoRA strengthens identity while the character-sheet LoRA keeps handling clothing/body.

**How to build a character sheet (for now):**take (or generate) a normal reference photo, then use ChatGPT (or another image-editing model, e.g. Nano-Banana) with this prompt:

“Convert the person in the photo into a character reference sheet with four panels arranged left to right: a close-up of the face, then full-body front, side, and back views, on a white background.”

Full write-up (longer prompt variant, gotchas) incharacter\_sheet\_prompt\.md.

🛠️ We’re training a dedicatedFlux Klein 9B LoRAto generate these character sheets directly (more consistent, no API needed) —**still in training, not released yet.**Until it’s out, use the prompt above with ChatGPT/gpt-image for the most faithful character sheet.

Usage difference vs. the close-up model:the character-sheet reference must be exactly1536×1024— that’s the fixed resolution it was trained at, independent of the output video’s resolution. When using theLTX Identity Transfernode with this checkpoint, set**ref\_resize\_mode = native\_resolution**(not the defaultmatch\_target) — otherwise the node resizes the sheet down to the video’s resolution and destroys the facial/clothing detail the model relies on. This setting doesn’t matter for the plain close-up reference, where ref and video share the same resolution bucket.

https://huggingface.co/Alissonerdx/LTX-Best-Face-ID#character-sheet-examplesCharacter-Sheet Examples

https://huggingface.co/Alissonerdx/LTX-Best-Face-ID#training-detailsTraining details

Base modelLTX-2 (22B)MethodLoRA (rank 128, alpha 128)Conditioningoverlap reference latent + TASS-RoPE source-phase (source\_id=2,phase\_scale=1\.0)Aux lossArcFace identity loss (+ temporal identity consistency)Datareference-image ↔ video pairs (OpenS2V subset + HuMoSet), close-up / frontal identitiesCharacter-sheet continuationsame recipe, continued from the close-up checkpoint on 4-panel character-sheet references (face + front/side/back)

https://huggingface.co/Alissonerdx/LTX-Best-Face-ID#reference-image–what-works-bestReference image — what works best

This model was trained on face-focused reference images. Looking at the training data, every reference follows the same pattern:

• Close-up / bust crop — roughly chest-up, with the face large and clearly visible.
• A single subject, centered in the frame.
• Frontal or near-frontal (slight 3/4 angles are fine, but the face is always well exposed).
• Tightly cropped on the face/upper body — the training crops are about 460×406 (nearly square).

Recommendation: for the best identity transfer, give the model a reference that looks like this — crop your image to the face / upper body. Full-body shots, wide shots, or images where the face is small and far away work noticeably worse, because the model never saw references like that during training.

In short: a clean, frontal, well-lit close-up of the face is the ideal reference.

**Using a full-body reference instead?**SeeCharacter-Sheet Referenceabove — the close-up guidance on this page is specifically for the base checkpoint.

https://huggingface.co/Alissonerdx/LTX-Best-Face-ID#captions-importantCaptions (IMPORTANT)

This model was trained on captions in the**ref\_t2v:format thatdescribe the scene and the person’s action**(not just “a person”). Identity in LTX is strongly prompt-driven — the more the prompt describes, the better the identity holds. For best results:

• Prefix prompts with**ref\_t2v:**.
• Describe theaction, setting, framing, camerain present-progressive.
• Describingidentity attributesof the reference person (skin tone, hair, eyes, facial hair, glasses, face shape)noticeably improvesthe result.

The shared ComfyUI workflow includes aPrompt Enhancerthat looks at the reference image (via the model’s own multimodal Gemma-3 text encoder) and automatically enriches your prompt with the person’s identity attributes — so you don’t have to describe them by hand.

Example (enhanced):

ref_t2v: A light-skinned man with long dark-brown hair past his shoulders, narrow rectangular
metal-frame glasses, light blue-gray eyes and light stubble is folding clothes in a laundry room,
medium-wide shot. He places the folded clothes into a basket, then carries it across the room.
Neutral indoor lighting, simple domestic environment.

https://huggingface.co/Alissonerdx/LTX-Best-Face-ID#usage-comfyuiUsage (ComfyUI)

Requires the BFS Nodes:https://github.com/alisson-anjos/ComfyUI-BFSNodes

1. Install ComfyUI-BFSNodes (ComfyUI Manager → Install via Git URL, or clone intocustom\_nodes/). Dependencies (insightface,transformers, …) install automatically.
2. Load LTX-2 (checkpoint + the Gemma-3 text encoder / CLIP) as usual.
3. Add theLTX Identity Transfer(overlap + source-phase) node; feed it the reference image. Using the character-sheet checkpoint? Set**ref\_resize\_mode = native\_resolution**(seeCharacter-Sheet Reference).
4. Loadthis LoRAon the MODEL path.
5. Use the shared workflow (it wires thePrompt Enhancer→ identity node → sampler).

A ready-madeworkflow JSONis included alongside the model.

https://huggingface.co/Alissonerdx/LTX-Best-Face-ID#about-the-arcface-projector-optional–you-dont-need-itAbout the ArcFace projector (optional — you don’t need it)

The identity node has an optionalArcFace projectordropdown (it projects a face embedding into extra conditioning tokens).You can leave it onNone.In our experiments the projector’s effect wasmarginal— identity is carried almost entirely by the overlap reference latent (and the prompt), so the projector adds very little on top and is not required for this model. It is kept only as an optional extra; skipping it keeps the graph simpler with no meaningful quality loss.

https://huggingface.co/Alissonerdx/LTX-Best-Face-ID#tips-for-best-resultsTips for best results

• Use aclean, frontal, well-lit close-upas the reference — identity quality tracks reference quality closely. Need clothing/body consistency too? Use thecharacter-sheet checkpointinstead of a plain close-up.
• Let thePrompt Enhanceradd the identity attributes (or write them yourself).
• Keep the reference framing simple; busy/occluded reference faces degrade identity.

https://huggingface.co/Alissonerdx/LTX-Best-Face-ID#limitationsLimitations

• **First-frame / “mask” tendency:**because identity comes from the raw referencelatent, the reference appearance (and framing / accessories like glasses) can partially copy into the video — it can look like the reference is “pasted”/masked rather than freshly generated. Using a clean frontal reference and the trim-first-frame node mitigates this.
• **Prompt dependence:**identity is much stronger when the prompt describes the person. A bare, generic prompt gives weaker identity — use the Prompt Enhancer.
• **Close-up bias (base checkpoint):**the base training data skews to close-up / frontal faces, so full-body or large-angle shots hold identity less wellwith the base checkpoint— use the character-sheet checkpoint for those cases instead.
• **Metric caveat:**ArcFace similarity is unreliable on small / turned / occluded faces — judge visually, not only by score.

https://huggingface.co/Alissonerdx/LTX-Best-Face-ID#citationCitation

If you use this model, please cite the base model and the technique:

@misc{ltx_best_face_id,
  title  = {LTX-Best-Face-ID: LTX-2 Identity LoRA (Reference-to-Video)},
  author = {Alissonerdx},
  year   = {2026},
  howpublished = {\url{https://huggingface.co/Alissonerdx/LTX-Best-Face-ID}}
}

• **Base model:**LTX-2 (Lightricks).
• TASS-RoPE / reference conditioning — ST-DRC:

@article{chen2026stdrc,
  title  = {Spatial-Temporal Decoupled Reference Conditioning for Identity-Preserving Text-to-Video Generation},
  author = {Chen, Yuheng and Hu, Teng and Wang, Yuji and He, Qingdong and Ma, Lizhuang and Zhang, Jiangning},
  journal = {arXiv preprint arXiv:2606.02441},
  year   = {2026}
}

• **Identity loss:**ArcFace (Deng et al., 2019) via InsightFacebuffalo\_l.
• Tooling:ComfyUI-BFSNodes.

https://huggingface.co/Alissonerdx/LTX-Best-Face-ID#acknowledgementsAcknowledgements

Trained on top of LTX-2. Identity supervision via InsightFace ArcFace. ComfyUI integration via the BFS Nodes.