Repository navigation

Add modular pipeline for HunyuanVideo 1.5 - #13389

Merged

yiyixuxu merged 19 commits into

huggingface:mainfrom

akshan-main:modular-hunyuan1.5

Apr 16, 2026

Commits on Apr 11, 2026

Add modular pipeline support for HunyuanVideo 1.5
akshan-main
committed
48b4f3f
Fix I2V latent/cond spatial dimension mismatch
akshan-main
committed
f029fc5
Fix guidance_scale default to 7.5 matching ClassifierFreeGuidance
akshan-main
committed
5cc2ed1
Fix tokenizer type: use Qwen2TokenizerFast to match model
akshan-main
committed
db41b41
Fix system message string formatting to match standard pipeline
akshan-main
committed
bfbb6b4
Rewrite HunyuanVideo 1.5 modular: use standard pipeline methods directly
akshan-main
committed
aec5b38
Remove I2V exports (T2V only for now)
akshan-main
committed
d9ad0ef
Fix encoder: use static methods directly instead of encode_prompt
akshan-main
committed
5923b5e
Inline all standard pipeline methods, remove runtime dependency
akshan-main
committed
1691e31
Add HunyuanVideo 1.5 image-to-video modular blocks
akshan-main
committed
d189ef1
Fix missing FrozenDict import in before_denoise.py
akshan-main
committed
cfd5de4
auto-generated docstrings via #auto_docstring
akshan-main
committed
acada9b
Fix ruff lint and format issues
akshan-main
committed
6ec688d
use InputParam/OutputParam templates and fix ruff
akshan-main
committed
93d0426
Address LTX review feedback here like add AutoBlocks, refactor I2V latents, lift encoders
akshan-main
committed
86fe24a
Add workflow map, workflow tests, auto docstrings, export only AutoBlocks
akshan-main
committed
77cd7d6

Commits on Apr 16, 2026

Merge branch 'main' into modular-hunyuan1.5
yiyixuxu
authored
458c97c
Address Claude CI review
akshan-main
committed
da4f4ef
Address claude CI review 2
akshan-main
committed
2bfc708