Add LTX 2.0 Video Pipelines #12915
dg845
merged
103 commits into
main
from
ltx-2-transformer
Initial LTX 2.0 transformer implementation
aa602ac4
Add tests for LTX 2 transformer model
b3096c3c
Get LTX 2 transformer tests working
980591de
Rename LTX 2 compile test class to have LTX2
e100b8f2
Remove RoPE debug print statements
780fb61d
Get LTX 2 transformer compile tests passing
5765759c
Fix LTX 2 transformer shape errors
aeecc4d7
Initial script to convert LTX 2 transformer to diffusers
a5f2d2da
Add more LTX 2 transformer audio arguments
d86f89dd
Allow LTX 2 transformer to be loaded from local path for conversion
57a8b9c3
Improve dummy inputs and add test for LTX 2 transformer consistency
a7bc052e
Fix LTX 2 transformer bugs so consistency test passes
bda3ff13
Initial implementation of LTX 2.0 video VAE
269cf7b4
Explicitly specify temporal and spatial VAE scale factors when conver…
baf23e2d
Add initial LTX 2.0 video VAE tests
5b950d6f
Add initial LTX 2.0 video VAE tests (part 2)
491aae08
Get diffusers implementation on par with official LTX 2.0 video VAE i…
a748975a
Initial LTX 2.0 vocoder implementation
c6a11a55
Merge pull request #3 from huggingface/ltx-2-vocoder
8bfeb4af
Merge pull request #2 from huggingface/ltx-2-video-vae
b1cf6ff8
Use RMSNorm implementation closer to original for LTX 2.0 video VAE
6c56954f
start audio decoder.
b34ddb17
init registration.
f4c2435d
up
e54cd6bb
simplify and clean up
907896d5
up
4904fd6f
Initial LTX 2.0 text encoder implementation
0028955c
Rough initial LTX 2.0 pipeline implementation
d0f9cdaa
up
5f0f2a03
up
58257eb0
up
059999a3
up
8134da6a
resolve conflicts.
409d651b
Merge pull request #5 from huggingface/audio-decoder
7bb4cf76
Add imports for LTX 2.0 Audio VAE
5f7e43d1
Conversion script for LTX 2.0 Audio VAE Decoder
d303e2a6
Merge branch 'ltx-2-transformer' into ltx-2-t2v-pipeline
ae3b6e7c
Add Audio VAE logic to T2V pipeline
54bfc5d6
Duplicate scheduler for audio latents
6e6ce205
Support num_videos_per_prompt for prompt embeddings
cbb10b8d
LTX 2.0 scheduler and full pipeline conversion
595f485a
Add script to test full LTX2Pipeline T2V inference
3bf73697
Fix pipeline return bugs
fa7d9f77
Add LTX 2 text encoder and vocoder to ltx2 subdirectory __init__
a56cf234
Fix more bugs in LTX2Pipeline.__call__
90edc6ab
Improve CPU offload support
1484c431
Fix pipeline audio VAE decoding dtype bug
f9b94765
Fix video shape error in full pipeline test script
e89d9c19
Get LTX 2 T2V pipeline to produce reasonable outputs
b5891b19
Merge pull request #4 from huggingface/ltx-2-t2v-pipeline
0c412974
Make LTX 2.0 scheduler more consistent with original code
581f21c4
Fix typo when applying scheduler fix in T2V inference script
e1f0b7e2
Refactor Audio VAE to be simpler and remove helpers (#7)
280e3478
Add support for I2V (#8)
46822c43
Merge branch 'ltx-2-transformer' into make-scheduler-consistent
6a236a27
Denormalize audio latents in I2V pipeline (analogous to T2V change) (…
bd607b97
test i2v.
d3f10fe5
Merge pull request #10 from huggingface/make-scheduler-consistent
aae70b90
Move Video and Audio Text Encoder Connectors to Transformer (#12)
caae1676
up (#19)
0be4f316
address initial feedback from lightricks team (#16)
c5b52d6c
When using split RoPE, make sure that the output dtype is same as inp…
2fa4f847
Fix apply split RoPE shape error when reshaping x to 4D
bff98911
Add export_utils file for exporting LTX 2.0 videos with audio
cb50cacb
Merge pull request #20 from huggingface/video-export-utils-file
ce9da5d4
Tests for T2V and I2V (#6)
93a417f2
resolve conflicts.
9b8788cc
up
c039c87b
use export util funcs.
550eca35
Point original checkpoint to LTX 2.0 official checkpoint
ef199118
Allow the I2V pipeline to accept image URLs
ace2ee93
make style and make quality
dd81242e
yiyixuxu
commented
on 2026-01-06
Merge branch 'main' into ltx-2-transformer
2fc57894
JoeGaffney
commented
on 2026-01-06
remove function map.
57ead0b5
remove args.
c39f1b87
update docs.
bdcf23ec
update doc entries.
61e0fb4b
disable ltx2_consistency test
8c5ab1fd
Merge branch 'main' into ltx-2-transformer
64b48c17
Simplify LTX 2 RoPE forward by removing coords is None logic
5e0cf2b2
make style and make quality
d01a242c
Support LTX 2.0 audio VAE encoder
79cf6d7b
sayakpaul
commented
on 2026-01-07
Merge branch 'main' into ltx-2-transformer
cc28cf76
resolve conflicts
91ee2dd2
Merge branch 'ltx-2-transformer' of github.com:huggingface/diffusers …
5269ee50
Apply suggestions from code review
a17f5cb6
Remove print statement in audio VAE
964f1068
up
4dfe5099
Merge branch 'main' into ltx-2-transformer
249ae1f8
Fix bug when calculating audio RoPE coords
040c1188
Ltx 2 latent upsample pipeline (#12922)
44925cb3
Fix latent upsampler filename in LTX 2 conversion script
5e500467
Add latent upsample pipeline to LTX 2 docs
2b85b93e
Add dummy objects for LTX 2 latent upsample pipeline
40ee3e33
Set default FPS to official LTX 2 ckpt default of 24.0
99ff722e
Set default CFG scale to official LTX 2 ckpt default of 4.0
165b9454
Update LTX 2 pipeline example docstrings
1a4ae58c
make style and make quality
b4d33df9
Remove LTX 2 test scripts
724afee9
yiyixuxu
approved these changes
on 2026-01-08
Fix LTX 2 upsample pipeline example docstring
d24faa71
Add logic to convert and save a LTX 2 upsampling pipeline
353f0dbd
Merge branch 'main' into ltx-2-transformer
0c9e4e22
Document LTX2VideoTransformer3DModel forward pass
f85b969a
Reviewers
Assignees
No one assigned
Labels
roadmap
Milestone
No milestone
Login to write a write a comment.
Login via GitHub