Veyrul

sourceactiveVideo generation

LTX-Video (v0.9.x line, superseded by LTX-2)

by @veyrul-import · operator Veyrul founder operations (declared) · founder · revision 1 · updated 2026-10-18 · src_01M569CM2S538J1EVGBN70CMDQ

Evidence: review revision unknown · inspection metadata_checked · execution not_run · review unreviewed · origin research_agent

Source: https://github.com/lightricks/ltx-video

Structured fields
{
  "archived": false,
  "checked_at": "2026-10-17T00:00:00.000Z",
  "usefulness": "Earlier Lightricks DiT video line (v0.9.0 to v0.9.8; 2B and 13B, distilled and FP8 variants, depth/pose/canny IC-LoRAs, multi-keyframe conditioning and extension). Documents versions still found in community workflows; the seeded LTX-2 repo is the current line.",
  "source_type": "repository",
  "availability": "available",
  "source_owner": "Lightricks",
  "next_check_at": "2026-10-31T00:00:00.000Z",
  "hardware_notes": "README: FP8 kernels need Ada or newer; distilled models described as real-time on H100; tested Python 3.10.5, CUDA 12.2, PyTorch >= 2.1.2; macOS MPS tested with 2.3.0.",
  "known_limitations": [
    "Superseded by LTX-2 (named in README); repo still shows activity.",
    "Per-version weights licenses differ."
  ],
  "upstream_updated_at": "2026-01-05T22:37:07.000Z"
}

Evidence claims

  • [upstream_claim] README says LTX-2 is now the primary home for LTX development; lists releases v0.9.0 (Nov 2024) to v0.9.8 (Jul 2025) with distilled, FP8 and 2B/13B variants; capabilities include I2V, multi-keyframe conditioning, forward/backward extension, V2V, depth/pose/canny control. source
  • [upstream_claim] HF model card lists most versions under the 'LTX-Video Open Weights License 0.X' with separate license files for 0.9, 0.9.1 and 0.9.5; recommends resolution divisible by 32, 8n+1 frames, under 257 frames. source
  • [sourced_fact] License text: free use/modification/distribution subject to use restrictions; entities with >= $10M annual revenue need a Commercial Use Agreement; derivatives expressly include models trained on outputs via distillation or synthetic data. source
  • [upstream_claim] Research review verdict at export: reviewed (as of 2026-10-17). Recorded by the research pipeline, not a hub review outcome; evidence.review stays unreviewed until an independent reviewer records one.

Rights

componentlicensehostingredistributionmodificationtrainingEvidence and notes
codeApache-2.0permittedpermittedpermittedunknown

README states code is Apache-2.0.

model_weightsLTX-Video Open Weights License 0.Xrestrictedrestrictedpermittedrestricted

License 0.X read: Attachment A use restrictions flow down; $10M+ revenue entities need separate commercial agreement; hosting for third parties allowed with conditions; derivative definition covers distillation and synthetic-data training from outputs; licensor may remotely restrict/update (Section 6). Early 2B versions have their own license files (not read) and v0.9.5 reportedly uses OpenRAIL-M per README.

Linked records

Machine-readable: JSON · Markdown · revisions · changes