aestx: add aestx-prev-single-ipadapter.json (drop-in UI + tested API versions)

aestx: add aestx-single-ipadapter.json (drop-in UI + tested API versions)

Upload 3 files

Remove paper folders — centralized in private aleph65/Papers

makeup-extract: install + run scripts for Makeup Extraction of 3D Representation (EG2023) (single photo -> FLAME UV materials + extracted makeup layers)

facelift: add export pass (2048px orbit RGBA renders + OpenCV cameras + run_meta, versioned output layout)

facelift: install + run scripts for FaceLift (single photo -> 3D Gaussian head)

download_missing_models: serial downloads by default, 10s stall watchdog with auto-restart (x3) then curl fallback, add --last flag

Add 3dgazenet: 3DGazeNet gaze estimation + dense eye mesh scripts (native env, like p3dm/ffhq-uv/deca) + stevie example outputs

Add deca: DECA detailed 3D face reconstruction scripts (native env, like p3dm/ffhq-uv) + stevie example outputs

Add ffhq-uv: FFHQ-UV RGB fitting scripts (native env, like p3dm) + stevie example outputs

p3dm: README.md - export raw fit data (frame ckpt, float npys, cropbox, seg, landmarks) by default; fix --force glob collision

p3dm: p3dm_run.sh - export raw fit data (frame ckpt, float npys, cropbox, seg, landmarks) by default; fix --force glob collision

p3dm: p3dm_install.sh - export raw fit data (frame ckpt, float npys, cropbox, seg, landmarks) by default; fix --force glob collision

p3dm: add README.md (pixel3dmm single-photo -> 3D head pipeline)

p3dm: add p3dm_run.sh (pixel3dmm single-photo -> 3D head pipeline)

p3dm: add p3dm_install.sh (pixel3dmm single-photo -> 3D head pipeline)

download_missing_models: fix mid-download deadlock (disable HF_XET_HIGH_PERFORMANCE/hf_transfer, add stall watchdog + resumable curl fallback)

serverless/client: CALLING-TIPS.md for downstream agents

serverless: qwen-edit-nsfw + faceswap-lissie endpoints (configs, clients, docs); handler fixes (passthrough LoadImage rewrite, seed range/map, output cleanup); .git stripped from node layers

serverless: client scripts (sync/async) + README; timings from live endpoint tests

README: point to serverless/

serverless: RunPod worker (handler, build/deploy scripts, Dockerfile, qwen-edit-turbo endpoint)

Add new-faceswap-lissie-v4 (klein distilled no-LoRA head swap, winner of faceswapnew experiments)

Add models for faceswapnew workflows: bf16 Qwen3-8B text encoder, FLUX.2 Klein base 9B, BFS Head V5 (2511)

Add faceswapnew-* head-swap experiment workflows (klein quickwins/nolora/base/res2s/cropstitch, qwen2511 bfsv5/lightning8) + research notes

Mirror segformer_face (FaceSegment dep for faceswap-v14)

Add new-krea2-faceswap-v14: v13 + FaceSegment face mask into ref_boost_mask

Add new-krea2-faceswap-v13 workflows: AR-matched output (<=1.85MP, /16 dims), grounding_px 1024, turbo 12 steps

Add new-krea2-faceswap-v12 workflows (benchmark-winning v1.2 face swap recipe: v1_2 LoRA + fit + ref_boost 4)

Add new-krea2 identity-edit workflows (std + max) from The AI Blueprint Krea 2 Identity LoRA video

Add text_encoders/qwen3vl_4b_bf16.safetensors (Krea 2 identity-edit video workflows)

Add loras/krea2_identity_edit_v1_2.safetensors (Krea 2 identity-edit video workflows)

Add diffusion_models/krea2_turbo_mxfp8.safetensors (Krea 2 identity-edit video workflows)

download_missing_models.sh: GITHUB_TOKEN auth + codeload tarball fallback for git clones (datacenter IPs get 403 from github.com)

ltx-2.3-i2v-nsfw-te API: DURATION_SECONDS constant + 10s clamp

ltx-2.3-i2v-nsfw-te: expose DURATION_SECONDS constant (node 340) with 10s ceiling clamp

ltx-2.3-i2v-bf16 API: DURATION_SECONDS constant + 10s clamp

ltx-2.3-i2v-bf16: expose DURATION_SECONDS constant (node 340) with 10s ceiling clamp

ltx-2.3-i2v API: DURATION_SECONDS constant + 10s clamp

ltx-2.3-i2v: expose DURATION_SECONDS constant (node 340) with 10s ceiling clamp

ltx-2.3-i2v-nsfw-te API: prompt_enhance off by default

ltx-2.3-i2v-nsfw-te: prompt_enhance off by default (text-only enhancer caused scene cuts in i2v)

ltx-2.3-i2v-bf16 API: prompt_enhance off by default

ltx-2.3-i2v-bf16: prompt_enhance off by default (text-only enhancer caused scene cuts in i2v)

ltx-2.3-i2v API: prompt_enhance off by default

ltx-2.3-i2v: prompt_enhance off by default (text-only enhancer caused scene cuts in i2v)

ltx-2.3-i2v-nsfw-te API: auto portrait/landscape detection + hard first-frame anchoring

ltx-2.3-i2v-nsfw-te: auto portrait/landscape detection + hard first-frame anchoring

ltx-2.3-i2v-bf16 API: auto portrait/landscape detection + hard first-frame anchoring