feat: expose Reference Audio in V3.5 sampler
This commit is contained in:
+39
-38
@@ -1,13 +1,13 @@
|
||||
dca65fb1393e67f59c6f81f9411e3e745eb8f1097e9a2adf7ffc6f4e3901ca9f __init__.py
|
||||
15b900d02624bea6cfe8ea25b22266a0c68b38b9f10c9089be0e38988bf2e34d .github/workflows/tests.yml
|
||||
0a2f69d4cec89d440f5d5aa1ec68a7c9a741382d313f80dcddf96d13f70b3e68 .gitignore
|
||||
16ea21aa8fe115ce8f6538a0268bc64762be2fd6931b3a0071a86a8cc9d218cf AGENTS.md
|
||||
1313e8e2301cfd6aff6c97bc50858bf9450210c83bd4166a74e61fd9c66c0498 CHANGELOG.md
|
||||
31c3aebb287fe73bfb567db41b4539058f8ef074dae400405f7b7a44dc5b9cbc .github/workflows/tests.yml
|
||||
85eada29c12706d55b78e4d8100899d1dadde4d9867406b8bac4eec22b81c6b6 .gitignore
|
||||
b3df39256d3d6a2a955c6447365714ff4460d461cf31a714e841f97d5a4c2353 AGENTS.md
|
||||
154b9568464a52226c3aa549b4b8f7eab09da3ed08f53e0eab3595584d11756f CHANGELOG.md
|
||||
e97e47c7585e81d9ec5fd32c0ca0b64699fd4610cd03e9d6575a91c30a10ca4e compatibility.py
|
||||
4d62508f900af9f93ec320afb900c27cc9a300d1e0c0a94714258498c5ede5fc conditioning.py
|
||||
16148b2175b371e082f2c623f6084bb45919d9eb5e4c478f882920c6761b5f6a conditioning.py
|
||||
71e3df269e226b1ecad597f9831c2a10b2215c8a3783652e9e9f536e99ce46c9 constants.py
|
||||
cf255b8a635471229aebadc926acfdcacb65963adbcb9e0f9505d04eada2c6b5 continuation.py
|
||||
f9375629fa1a2c4d3538442f0c3192d7ba2f46fe9544fb2e21f0c144b53d608c docs/AB_TEST_LOG_JA.md
|
||||
23db6dadd0848345c74d066d421bdc6f695cd524cb6373078bcf105a10bb3830 docs/AB_TEST_LOG_JA.md
|
||||
c93a3e52e4deabe5ef0db7f34a7605856a543089574ba17227e697e9a2c0139a docs/ARCHITECTURE.md
|
||||
7660570840229b6fcbe62f0fbb52966d0099d64d7e6f01a739fbf9899293c8e4 docs/FUTURE_ROADMAP.md
|
||||
9a634b4ff6c9ca98285f6b250402a3d0e67d29f031e134d9bfcab54c119ed337 docs/images/comfyui-manager-h3-continuum.png
|
||||
@@ -27,8 +27,8 @@ ff8d901d05aa89f9b698c60b3d11534ef97982b3e22e9ede6255a85149a669c4 docs/images/v3
|
||||
a53796f7a904d675e80a2276c20d68502f03a3958529278e6ce9465baaf1c4af docs/images/v35-hires-fix-node.png
|
||||
5a23f232a8b795226e8255e6e2e3248b0130f1c7ce276d50370d0333d6a62ef0 docs/images/v35-second-pass-node.png
|
||||
93f07aa727b34802bce3d333426b8ba2af522500882a6d222d00cf577ca82e49 docs/SEAM_TEST_SOURCE_JA.md
|
||||
2c13aefd137db92a3dbf5576e9442d58012b1f0de1240c07d039ceeb702e7c1f docs/V2_1_4_UI_SIMPLIFICATION.md
|
||||
b08022df068576c759c8bf80ac6b6519693f3c65c917f1822aeb06ea359aa5b1 docs/V2_1_5_UI_SERIALIZATION_STABILITY.md
|
||||
90acddd795d84c5a24f79598670aec1655eaf58d45fd6d362dc758e2b880a071 docs/V2_1_4_UI_SIMPLIFICATION.md
|
||||
61805ebbfa2bb2451c4fc8a2bed046073cc135b93ddc4b52dca61a0459c8d3d6 docs/V2_1_5_UI_SERIALIZATION_STABILITY.md
|
||||
a032738b0bc5acb136e92d8bf7694643a5f518b9c7fc0fc878d66d81ba8b1b04 docs/V2_1_6_UNIFIED_UI.md
|
||||
384e3841603f7c6c8286ef147e93aaf7d056a286503a2a4fc21f397e4dec4876 docs/V34_DRIVING_AUDIO_VALIDATION.md
|
||||
1840c81b4090ff3c86c0c1b0acc2cc10d3d9839e383c81f262a5c59a67504cd6 docs/V35_HIRES_FIX.md
|
||||
@@ -37,7 +37,7 @@ f0806bb2b999794ddda072e98424dbd1b613df1cd8a096b12d979f124e066d8a docs/VALIDATIO
|
||||
cedd7b07901d6065079ef37380f0062d535a39c98c8f614e1b26582f549d7183 driving_audio.py
|
||||
6771c9905d35666121c4892d2ec26202505f96d244ce814ad9a87a02878952cd examples/H3_CONTINUUM_V1_WIRING.md
|
||||
f85ba3c69a473f0ac4bb96895c6d6532e8270121da65f858e364b19d31f05ef5 examples/H3_Continuum_V2_3x5s.json
|
||||
be419d6c773784823af7b58a3423ef2e21d3cc3e23c40e60e274560895ca034b examples/H3_Continuum_V2_README.txt
|
||||
9d88c187e80ab726afb4dc685dc2ac20039a2c6933f118caa23e2ba8e1c640cd examples/H3_Continuum_V2_README.txt
|
||||
4b94f779a98bfc555f6999113a694da07746cc3cf400bfc70ebc580592034d52 examples/H3_Continuum_V2_reference.webp
|
||||
c3cefcdf50df9d6609f583a008453b587c011a0fb41e290d5f1196ce62f58877 examples/H3_CONTINUUM_V2_WIRING_JA.md
|
||||
20f4e7c4ad085791f8b578f575d03e52ef09d0d4550aa58ad086c9857b59bf29 examples/H3_Continuum_V2.json
|
||||
@@ -46,36 +46,36 @@ f67febfdd7b2bfca2ad4e477c87a9a74dbe636231f9db3e1158b944d7b9f8397 examples/H3_Co
|
||||
5a26408c770381cdb259cb710de7dab12d7ddecca262bdbf35ba1566056050f5 examples/recommended_settings_v2.json
|
||||
b8275c8cecde12aee0418cea456d1e7f4b875cbb1bce32f1f3d9e06a44dd0be1 examples/v2_prompt_list_example.txt
|
||||
0e59ea821081cb93165aa3f88c6949736dd225837fe1cdcf1173aa5cd54c6d29 examples/v2_prompt_timeline_example.txt
|
||||
55bc60447d6cd783b9df049d04b6cebb6e4a633981434ac2b28e5a1527d4f4b4 examples/workflows/MiniMax_H3_Continuum_V32_turbo.json
|
||||
4e410aed8d7f6ebf11c98d4f71571e46855e4619c2bcbec921df70f29189b6b1 examples/workflows/MiniMax_H3_Continuum_V32.json
|
||||
55bc60447d6cd783b9df049d04b6cebb6e4a633981434ac2b28e5a1527d4f4b4 examples/workflows/MiniMax_H3_Continuum_V33_turbo.json
|
||||
4e410aed8d7f6ebf11c98d4f71571e46855e4619c2bcbec921df70f29189b6b1 examples/workflows/MiniMax_H3_Continuum_V33.json
|
||||
c3a6345242c598caa64fafe68a5e304ede59ffe3092ee70a285995110253ad65 examples/workflows/MiniMax_H3_Continuum_V32_turbo.json
|
||||
f69c9e9075bc3302dd91b1890b79e0e4d8f7b34c46da999bb51d92f81743664c examples/workflows/MiniMax_H3_Continuum_V32.json
|
||||
c3a6345242c598caa64fafe68a5e304ede59ffe3092ee70a285995110253ad65 examples/workflows/MiniMax_H3_Continuum_V33_turbo.json
|
||||
f69c9e9075bc3302dd91b1890b79e0e4d8f7b34c46da999bb51d92f81743664c examples/workflows/MiniMax_H3_Continuum_V33.json
|
||||
5eab33c110e07ec08ac66f289f20c295a90ca026ea5231ad6121755a28a2ded8 examples/workflows/MiniMax_H3_Continuum_V34_turbo.json
|
||||
660e955c983e41947f5abfd89e66fbe5759ffda01e6c1a7164d580b0c41b2120 examples/workflows/MiniMax_H3_Continuum_V34.json
|
||||
ed31ddd3d0af5e6f8dbd609e5ef51872322dbe52eaad8cae35e16342ac53dbfa examples/workflows/MiniMax_H3_Continuum_V35.json
|
||||
e8f41dee2240f1fcc6e8fd12448856513483ff3ae0427b2c554bb351f9ae8047 graph_contract.py
|
||||
3d14acac341c63207995e2bd45cd407fdb3e2308273b1c90605adfe6f2e0bd8f hardening.py
|
||||
79bd878b1c639c8ad77afe972a20810d1060f7903f0ae8ddc31ba22d84777ba1 hardening.py
|
||||
6abdbc29472c7f9d447085b189fc47b69e110b94ef83192f9689d04000408843 INSTALL_JA.txt
|
||||
23bddc5a41d2e997f4fbf789c7128e32f4f197f1405b2aef1c76794f57482446 install_windows.bat
|
||||
d74073e7ee292ebeba913af4288284f9f197cd91fceed6c62bf6c66b784c7cec install_windows.ps1
|
||||
6aa3f5780baa7bb0c0672b5f01a2f77e5f209da93f6373468cb274ea2f643d4b layout_adapter.py
|
||||
c44abbd789091bbead01250dbd5453e60539d2665c69f162335fdbd2936019c2 LICENSE
|
||||
15fd1cdc62ffa959b49fd5088bdb1d4b5fa406fa8e420de64603e86458a1a3b8 LICENSE
|
||||
d3e213005670e94dee33a240349a4d245cf9a3b6602099f11ed30cf5a32e2c21 media.py
|
||||
84cc75b4583a2851e657f3142a1ec4e099091cb7a960be7355e376269e4d3048 metadata.ini
|
||||
6d9527ba9cf7fa411f46056c93ef662a1800fc0fc7bd2b7a12b9f72e9f81b892 metadata.ini
|
||||
ef705a476e3f5bbd05559e7a0e7d3dfb9275badef0f8f65fb1a36ce27fbcb537 model_patch.py
|
||||
01ea71240361725972e41746f5a1c4ff90653e6ccc673085f46f427fbf65e4be nodes.py
|
||||
a230cd2f772021a683432e41f7028f6ea628e34dd19f617a3f98c174fdc97d58 PACKAGE_INFO_JA.txt
|
||||
a7084bc1d12ab25a1ebe29a089dba72d6c1147e0f5556837e6321277562ad1f5 PACKAGE_VALIDATION.txt
|
||||
59ac121d919a0c372979e4e7d6c8f55b28968d99f44ef26f2dc4490ea172dd0f PACKAGE_VALIDATION.txt
|
||||
b470819f16fc179654dc1600721a63f0a5a14e777f4c004d574b499da6886ba5 public_api.py
|
||||
16bc48b9ef6127af7e0e3ddd96791640cedf5159cd7c06452d72604f41998a52 pyproject.toml
|
||||
fe65a21c645b62ce3c03a80ef9146a36e5e30c4d3a715cd4842c2fd1bfb9d95f pyproject.toml
|
||||
94fdc2d545b2d4077a4d70229fd22ba3ccac21d49d1097d2873f76d04e9a3c7a pytest.ini
|
||||
3a3ab9f5b9860c4b5478e37e49c06b1b102d49a408c712a25e439b6a557d7071 README_JA.md
|
||||
dfcaceb3a9ed4784ca0017f864011cc9904c15dc95d5b59dd39d8a429aee28bb README.md
|
||||
e8d655533351f66268561cd215dff67c00cfab789bb6cf7d4115424c39dfb6d0 README_JA.md
|
||||
bbb55aec2b7bbbae3ae4a77bffd8defd8966dabecba456930602392504363495 README.md
|
||||
5d89c366b4f076fa73a2695557e0af749bf95df1dad0d55dbb2fe5b1c2519ad7 reference_audio.py
|
||||
8f1744dc99f1a3a222b7d6e986f6dac1edcc0abbbe8b61a1d2606b1ca3569b3d reference_video.py
|
||||
2995924fcb7bac27b548b01187d8276d2ab103c6f43416cf85400de3846c6297 reference.py
|
||||
f45e2cd438125bbf4730d809118c3c007f742cc719cc458178275da8f6759fca reference.py
|
||||
467edb2c188d024c2392aafe7423b3f835e16b10f570da76e2f7dda8a962c1c5 requirements.txt
|
||||
e02b523b5b1c1272c912c16d547050cc9bb9feb6c348743015bb5d059a4ead06 run_storage.py
|
||||
1c2b36c532552c71712a645a7c8524f09b3afe4179ec2cfbf69f853f1ac2952d run_storage.py
|
||||
6056de6c0a5322a8a51ba5a114538dca2ef67788cd1afba10fedd483b95f6afa state_io.py
|
||||
3a359a0467725cc292ccb4c86cf0c3d30bb6a114d01f5345792b09ea3d45441d state.py
|
||||
43ec4023dc06c941e5d7a04db4b8a89fb58d744ceb986f6eb9bf534d4fdcb87b temporal.py
|
||||
@@ -100,19 +100,20 @@ d4c53d2c260b635d70fa211029d608f63776e0a56c20253475fa47c58363f8a7 tests/test_sta
|
||||
912661d91c7f11f5fcaf0bd14d7d3ecf5c74c33b1507272f0ae72af6e5bffa6f tests/test_v2_decoder.py
|
||||
36084a4f37e56786fa82200dad3e13a6a1a1ddde4f4eb5654117a88ef2659b99 tests/test_v2_prompts.py
|
||||
5dfdd4b108ba7444c0f5f371c3b1e244b5735d23136127b9818fd60738c78b68 tests/test_v2_registration.py
|
||||
0a520b34b12e1d2556d50480c43318de7eab4221afb96ef4760127c8bc90c890 tests/test_v2_seam.py
|
||||
273188a69b3d39eb7e190c16c593db440c5427e51435ddfbfe4e1b71f00b414c tests/test_v2_seam.py
|
||||
74e623120a963cb96b68651b44f47a836ea8449d30ce89d749c71ffad4766d4e tests/test_v2_seeds_motion.py
|
||||
b42bcc0983147462d96c6b71cc4a38e846995dafa6eebefc3ae0d9bb36caf359 tests/test_v2_sequence_orchestration.py
|
||||
0b5b20a8318862d167f135720e0489b8d4f93406d67c6646e052e35b03dcb876 tests/test_v2_session.py
|
||||
58438c3ed814adf40cefde098a37cae00ddcb703b7e5548bbbb295e54bed5a6d tests/test_v2_widget_migration.py
|
||||
0410e2d10af76a83eee0e622d63b8d91662d9516c990aa657a15066e30604f58 tests/test_v217_stable_facade.py
|
||||
78bc478a5e4edda902c80849c911782a99e75c06dfb04127391b6cc5b27b28e1 tests/test_v3_hardening.py
|
||||
4e777a8c6456328a96f1647033e1d74e3791559b21852b2523c69ed46acc7ff7 tests/test_v217_stable_facade.py
|
||||
7c81dd04af121e07fd63bac8e9f83c328219dc95c4939c021d75927c47cd3458 tests/test_v3_hardening.py
|
||||
eb5b27f8f0d79529c56ef344a5601703642c3cba5388c9768116223e2ed58a36 tests/test_v3_video_seam.py
|
||||
eec67e2daddbbe4f5933a3ac0dd9c3188f750081d377b7d08fd6d5a02ff4a529 tests/test_v31_run_storage.py
|
||||
3af78ab85c8e19dee8d0a98bfd11a2c0e88c30d87256da1f9699c49a19a684e9 tests/test_v32_reference.py
|
||||
8747b9abfa58a1bdc2381c9c86628c2c44321071a79de63a5a46a13c7c5deeb8 tests/test_v31_run_storage.py
|
||||
d66489f9ead30686ee82a4e1f24a7d7c4a36f835678812e1fca58f3bbf46ecaf tests/test_v32_reference.py
|
||||
f692476d0a7795b36bc5d9b97c10b7c6ba9686122b8f7987629948141e5dbab3 tests/test_v34_driving_audio.py
|
||||
7fd13dd3e969b623f525e38753252b179d55988b8fb794326137899d1fe7c6b6 tests/test_v351_reference_audio.py
|
||||
8b75c23dccfd1ac302596187f2fbbd1c9b71fddcb5e666aa4f9eedfd8b5b4b93 tests/test_v35_compact_ui_frontend.py
|
||||
c5a122cb9510320820f6057fc29e6bef4b42e19f141d4d9090c817702e7d151e tests/test_v35_context_sampler.py
|
||||
b7267778fd1e77bd12d7c6e71c200fd0a4367cd8ae3276823ddeba16e015cfa6 tests/test_v35_context_sampler.py
|
||||
c97d1323fa3d66c28dd1974340dbc17a3e6d5a79d4a5206b82d48bfb1a14b4ce tests/test_v35_hires_fix.py
|
||||
9ceac8c3c6e4e57ec8b53b48f3b547c204df36a3f6281328aa9ab5ebc2e6a0b6 tests/test_v35_latent_resize.py
|
||||
59a07f2172014dfdc0a447775324d7cd049f20b823276dfaf9b7b3f79628f5a8 tests/test_v35_pixel_vae_resize.py
|
||||
@@ -128,33 +129,33 @@ ffd61f3f31c9b6430e1169007ee906c8f5000eba63ee44e2976ad4f5f1112671 v2/__init__.py
|
||||
2003902c87bba09cfa143f5a9e4248f3a78c0fb544449a75fd40ba2cacb35655 v2/context_diagnostics.py
|
||||
cde27db19b677e318cc4dfb4235be7a11429c19945f4083052a91fe117e74c82 v2/decoder.py
|
||||
31012ec48a64c7467ad869c79982b7f8502c1785c33f51ee1b5727ce8b4243fb v2/diagnostics.py
|
||||
87b054be8754ebfcd85e106a7b221604611d3bf05de03380565ee3562d1c823b v2/h3_builder.py
|
||||
20b8132fb231c5b8ce61da2c1fbad310c22f332d66926a706523009deed54f65 v2/h3_builder.py
|
||||
aea01a48a8c03a25ed97ea05e5b137f89e388832d1be7d38f2da2be3d1b9830b v2/motion.py
|
||||
080e9d597e2586eb88c448c7db32b9e3661f5d21a4c270f163696966ffa6c7e0 v2/nodes.py
|
||||
a8432732548ae9521ee445a017e1d6d7f65b0e7bb120a89d6cfe1c18834f377c v2/nodes.py
|
||||
3c2145d47924b53f8f6ec50806f1b7113120ac314142d8b7809c4f2064ad6aef v2/prompts.py
|
||||
4080f1875b958b316e8e76852c4be5527a6de5c71d604e7642d3098478eb4a36 v2/sampling.py
|
||||
2fd811794c4cc0f2ce050250b80484876454926b9f20fce6af3a97d728e5acdf v2/seam_guard.py
|
||||
be0d8585fa57e0ccd6d02577bbc7299629e5bfe20e186747f571fccc98b67d7e v2/seam_types.py
|
||||
12cfcc0c71ca087fe39a5ce2922959a62b2bb311bfe664f05c4480affce03708 v2/seeds.py
|
||||
e6c15105fe66b9d7786602c8e6d51dd256428d80635cee686b0e35d66ab3b794 v2/sequence.py
|
||||
e1fa369acf61e17d25e7dfe24eb84a4a44dda1622f85a5f2e2b04210f1f38c85 v2/sequence.py
|
||||
13213e7709fd02bb810af3c18e4cd7d010f8cdfb121a4b99866a9f2333331089 v2/session_io.py
|
||||
3ea23dbce6fcd217b6e81235855479b9b16fed7a21693cf156f9014555c4fe13 v2/session.py
|
||||
78c830cdf49034c6e3b2cd3da5914d5cfc5638eb63ec6dbd1bab13c50d6d4caf v3/__init__.py
|
||||
b3dc61503d35d9ef0a0dc7fed1c9b2fdc00fcb2072d4e39f64ae0826f9ec7dca v3/assembly_v35.py
|
||||
7815230e0a9540bc2edeaf818e940b7a1fdd817fd217ba8a4f9cb21aaf6a308e v3/assembly.py
|
||||
1465a4cd53e4a90788c7cdbbbef5f7bc0dee2a8be04cc4c707863837077d9a47 v3/driving_nodes.py
|
||||
30e5ef19e182a288b21b76f644188cdb0763c367c8f468fad056bae65bc32f89 v3/assembly.py
|
||||
5af29b8247319dbb5321c711318ead2e75baf12bd26d1460bb8a4c58f8e631c8 v3/driving_nodes.py
|
||||
6cc7ed74666084c55beb9075304e92a9defafd81b9c6678239ca3859a08cddd7 v3/file_backed_buffer.py
|
||||
897343a4845f4ed59828bbfcfb16a22fcfce2c22907548bb3234c98dabcefff6 v3/hires_fix_nodes.py
|
||||
0858487e2b8c0060a856273ffc65d23165b2f253d183f506e27e419aea0da561 v3/latent_resize_nodes.py
|
||||
45bd1945b984b56369234053be2edb8455d4cff7bf976f7d9c7bef77a762b7eb v3/memory_attribution.py
|
||||
f8708118f70fa2aae22951e025ed3b9499951088988c817c28f6ef7f388f2570 v3/nodes.py
|
||||
5bfa482eaae3fdc04601c6bd6ea41c175632b581f7a3e0dc80ff408cae275b0f v3/nodes.py
|
||||
b7c4a35a5e73f528492bc6946b97b687c412f1d3f5f976c19d45c6dc20288c9a v3/pixel_vae_resize.py
|
||||
34ea38c03b6322288d88fc8595e3655a7a7afddb51709142a954eb7349f71ac2 v3/plan.py
|
||||
60ac543e3a5ae9029ff2eeb5c91b72280869acdf3ee4808676650f213b3438c6 v3/plan.py
|
||||
10eaab14eb8a70dea3345bf8bb20069d78ec5464aef5542a9a80f1ccfdeeb9b0 v3/refine_context.py
|
||||
519586b1b3f48c73d329f7a2aaa8a1bfc6fe31ac2ecb64f2b1fdf9c19026001a v3/refine_sampling.py
|
||||
da5edf3b4392c41e12ef73a0cd21799be42a327b18bd5a6989447b75163dcab1 v3/second_pass_nodes.py
|
||||
998d002c73c59ce88de1cf72afbb5f481a57cc31d365ba2fb4889ebf405dedc8 v3/second_pass.py
|
||||
4178dcfa442abdb06ce276e05bd7bcbaa6da28469ba5e57d3da3c663de4e4a9d v3/video_seam_v35.py
|
||||
2f07ace1462644f0bee553827b3571535de4d0c2d9ef847bd9e0ca73a4eed5cb v3/video_seam.py
|
||||
343bf84d768b93a59c8a418635bcad4133961431c9c6f7975d7d65b0fac25eb4 version.py
|
||||
ea79af984eb7d615f0d79384c8b505442af2b00cf7babfd53e82da2733927bcc web/project_id.js
|
||||
f8824d6c9280b0af7723722e06a78c2ce84faa1da1a9ee2e41ee4b03cfda8d37 v3/video_seam.py
|
||||
52173490b708846a794b625befde12952daf8aa070f762b2713aa8bfe660ebc8 version.py
|
||||
760f4a77128b80107477a475d2ae0e5440475ff3095d9653ad7249c9a534daaa web/project_id.js
|
||||
|
||||
@@ -277,6 +277,15 @@ class _VideoVAE:
|
||||
self.upscale_ratio = 16
|
||||
|
||||
|
||||
class _AudioVAE:
|
||||
def __init__(self):
|
||||
self.first_stage_model = torch.nn.Linear(2, 2)
|
||||
self.patcher = _Patcher(self.first_stage_model)
|
||||
self.vae_dtype = torch.float32
|
||||
self.audio_sample_rate = 32000
|
||||
self.latent_channels = 32
|
||||
|
||||
|
||||
def _sampling_contract(
|
||||
*, model=None, clip=None, video_vae=None, sampler=None, reroll_nonce=0,
|
||||
conditioning_mode="i2va", first_frame_hash="a" * 64,
|
||||
@@ -284,6 +293,8 @@ def _sampling_contract(
|
||||
upstream_graph_contract=None, upstream_graph_safe=None,
|
||||
upstream_graph_reasons=None,
|
||||
execution_semantics=None,
|
||||
reference_audio_contract=None,
|
||||
reference_audio_vae=None,
|
||||
):
|
||||
return build_sampling_contract(
|
||||
model=model or _Model(),
|
||||
@@ -314,9 +325,63 @@ def _sampling_contract(
|
||||
upstream_graph_safe=upstream_graph_safe,
|
||||
upstream_graph_reasons=upstream_graph_reasons,
|
||||
execution_semantics=execution_semantics,
|
||||
reference_audio_contract=reference_audio_contract,
|
||||
reference_audio_vae=reference_audio_vae,
|
||||
)
|
||||
|
||||
|
||||
def test_reference_audio_revision_separates_changes_and_restores_on_disconnect():
|
||||
model = _Model()
|
||||
clip = _Clip()
|
||||
sampler = _Sampler()
|
||||
audio_vae = _AudioVAE()
|
||||
common = {
|
||||
"model": model,
|
||||
"clip": clip,
|
||||
"sampler": sampler,
|
||||
"conditioning_mode": "t2va",
|
||||
"first_frame_hash": "",
|
||||
}
|
||||
disconnected, safe, reasons = _sampling_contract(
|
||||
**common,
|
||||
reference_audio_vae=audio_vae,
|
||||
)
|
||||
assert safe, reasons
|
||||
assert "reference_audio" not in disconnected["global"]
|
||||
assert "reference_audio_vae" not in disconnected["global"]
|
||||
|
||||
first_contract = {
|
||||
"reference_audio_contract_version": 1,
|
||||
"source_sha256": "a" * 64,
|
||||
"resolved_vae_sample_rate": 32000,
|
||||
}
|
||||
first, safe, reasons = _sampling_contract(
|
||||
**common,
|
||||
reference_audio_contract=first_contract,
|
||||
reference_audio_vae=audio_vae,
|
||||
)
|
||||
assert safe, reasons
|
||||
assert first["global"]["reference_audio"] == first_contract
|
||||
assert "reference_audio_vae" in first["global"]
|
||||
|
||||
changed_contract = dict(first_contract, source_sha256="b" * 64)
|
||||
changed, safe, reasons = _sampling_contract(
|
||||
**common,
|
||||
reference_audio_contract=changed_contract,
|
||||
reference_audio_vae=audio_vae,
|
||||
)
|
||||
assert safe, reasons
|
||||
restored, safe, reasons = _sampling_contract(
|
||||
**common,
|
||||
reference_audio_vae=audio_vae,
|
||||
)
|
||||
assert safe, reasons
|
||||
|
||||
assert revision_identity(disconnected) != revision_identity(first)
|
||||
assert revision_identity(first) != revision_identity(changed)
|
||||
assert revision_identity(restored) == revision_identity(disconnected)
|
||||
|
||||
|
||||
def test_terminal_seed_v2_revision_is_scoped_and_distinct():
|
||||
old_contract, old_safe, old_reasons = _sampling_contract(
|
||||
conditioning_mode="fl2va",
|
||||
|
||||
@@ -0,0 +1,154 @@
|
||||
from __future__ import annotations
|
||||
|
||||
import inspect
|
||||
import json
|
||||
from pathlib import Path
|
||||
|
||||
import pytest
|
||||
import torch
|
||||
|
||||
from ComfyUI_H3_Continuum_Join.reference_audio import (
|
||||
ReferenceAudioError,
|
||||
prepare_reference_audio_source,
|
||||
)
|
||||
from ComfyUI_H3_Continuum_Join.run_storage import automatic_project_key
|
||||
from ComfyUI_H3_Continuum_Join.v3 import driving_nodes
|
||||
from ComfyUI_H3_Continuum_Join.v3.driving_nodes import (
|
||||
H3ContinuumSamplerV34,
|
||||
H3ContinuumSamplerV35,
|
||||
)
|
||||
from ComfyUI_H3_Continuum_Join.v3.nodes import H3ContinuumSamplerProduction
|
||||
|
||||
|
||||
ROOT = Path(__file__).resolve().parents[1]
|
||||
|
||||
|
||||
def test_v35_appends_reference_audio_sockets_without_changing_v34_schema():
|
||||
v34_optional = H3ContinuumSamplerV34.INPUT_TYPES()["optional"]
|
||||
v35_optional = H3ContinuumSamplerV35.INPUT_TYPES()["optional"]
|
||||
|
||||
assert "reference_audio_1" not in v34_optional
|
||||
assert "reference_audio_vae" not in v34_optional
|
||||
assert list(v35_optional)[:-2] == list(v34_optional)
|
||||
assert list(v35_optional)[-2:] == ["reference_audio_1", "reference_audio_vae"]
|
||||
assert v35_optional["reference_audio_1"][0] == "AUDIO"
|
||||
assert v35_optional["reference_audio_1"][1]["display_name"] == (
|
||||
"Reference Audio"
|
||||
)
|
||||
assert v35_optional["reference_audio_vae"][0] == "VAE"
|
||||
assert v35_optional["reference_audio_vae"][1]["display_name"] == (
|
||||
"Reference Audio VAE"
|
||||
)
|
||||
assert "Driving Audio" in v35_optional["reference_audio_1"][1]["tooltip"]
|
||||
|
||||
|
||||
def test_v34_internal_reference_audio_parameters_are_appended_and_forwarded(
|
||||
monkeypatch,
|
||||
):
|
||||
parameters = list(inspect.signature(H3ContinuumSamplerV34.run).parameters)
|
||||
assert parameters[-4:] == [
|
||||
"capture_refine_context",
|
||||
"reference_audio_1",
|
||||
"reference_audio_vae",
|
||||
"kwargs",
|
||||
]
|
||||
|
||||
captured = {}
|
||||
reference_audio = object()
|
||||
reference_audio_vae = object()
|
||||
|
||||
monkeypatch.setattr(
|
||||
driving_nodes,
|
||||
"prepare_driving_audio_source",
|
||||
lambda *_args, **_kwargs: None,
|
||||
)
|
||||
monkeypatch.setattr(
|
||||
"ComfyUI_H3_Continuum_Join.reference_video.prepare_reference_video_source",
|
||||
lambda *_args, **_kwargs: None,
|
||||
)
|
||||
|
||||
def production_run(_self, **kwargs):
|
||||
captured.update(kwargs)
|
||||
return [], [], {}, "status"
|
||||
|
||||
monkeypatch.setattr(H3ContinuumSamplerProduction, "run", production_run)
|
||||
outputs = H3ContinuumSamplerV34().run(
|
||||
reference_audio_1=reference_audio,
|
||||
reference_audio_vae=reference_audio_vae,
|
||||
chunks=1,
|
||||
chunk_seconds=5.0,
|
||||
width=640,
|
||||
height=640,
|
||||
)
|
||||
|
||||
assert captured["reference_audio_1"] is reference_audio
|
||||
assert captured["reference_audio_vae"] is reference_audio_vae
|
||||
assert outputs[-1] is None
|
||||
|
||||
|
||||
def test_v35_run_forwards_reference_audio_without_duplicate_preparation(monkeypatch):
|
||||
captured = {}
|
||||
reference_audio = object()
|
||||
reference_audio_vae = object()
|
||||
|
||||
def v34_run(_self, *args, **kwargs):
|
||||
captured.update(kwargs)
|
||||
return [], [], {}, "status", None, {"groups": ()}
|
||||
|
||||
monkeypatch.setattr(H3ContinuumSamplerV34, "run", v34_run)
|
||||
H3ContinuumSamplerV35().run(
|
||||
reference_audio_1=reference_audio,
|
||||
reference_audio_vae=reference_audio_vae,
|
||||
)
|
||||
|
||||
assert captured["reference_audio_1"] is reference_audio
|
||||
assert captured["reference_audio_vae"] is reference_audio_vae
|
||||
assert captured["capture_refine_context"] is True
|
||||
assert captured["memory_attribution"] is True
|
||||
|
||||
|
||||
def test_reference_audio_vae_is_ignored_without_audio_and_required_with_audio():
|
||||
vae = object()
|
||||
assert prepare_reference_audio_source(None, vae) is None
|
||||
audio = {
|
||||
"waveform": torch.zeros((1, 2, 32000), dtype=torch.float32),
|
||||
"sample_rate": 32000,
|
||||
}
|
||||
with pytest.raises(ReferenceAudioError, match="reference_audio_vae is required"):
|
||||
prepare_reference_audio_source(audio, None)
|
||||
|
||||
|
||||
def test_unused_reference_audio_vae_does_not_change_automatic_project_key():
|
||||
base = {
|
||||
"1": {"class_type": "H3ContinuumSamplerV35", "inputs": {}},
|
||||
"2": {"class_type": "VAELoader", "inputs": {}},
|
||||
}
|
||||
vae_only = {
|
||||
**base,
|
||||
"1": {
|
||||
"class_type": "H3ContinuumSamplerV35",
|
||||
"inputs": {"reference_audio_vae": ["2", 0]},
|
||||
},
|
||||
}
|
||||
assert automatic_project_key(base, "1") == automatic_project_key(vae_only, "1")
|
||||
|
||||
|
||||
def test_published_v35_template_keeps_existing_sampler_socket_routes():
|
||||
template = ROOT / "examples/workflows/MiniMax_H3_Continuum_V35.json"
|
||||
if not template.exists():
|
||||
pytest.skip("published workflow template is not installed in this runtime copy")
|
||||
workflow = json.loads(
|
||||
template.read_text(encoding="utf-8")
|
||||
)
|
||||
sampler = next(
|
||||
node for node in workflow["nodes"] if node["type"] == "H3ContinuumSamplerV35"
|
||||
)
|
||||
routes = {item["name"]: item.get("link") for item in sampler["inputs"]}
|
||||
assert routes["reference_image_1"] == 321
|
||||
assert routes["reference_image_2"] == 320
|
||||
assert routes["reference_image_3"] == 318
|
||||
assert routes["reference_video_1"] == 322
|
||||
assert routes["driving_audio"] == 294
|
||||
assert routes["audio_vae"] == 295
|
||||
assert "reference_audio_1" not in routes
|
||||
assert "reference_audio_vae" not in routes
|
||||
@@ -69,6 +69,7 @@ def _capture_sequence(
|
||||
monkeypatch,
|
||||
*,
|
||||
terminal: bool,
|
||||
reference_audio: bool = False,
|
||||
session=None,
|
||||
diagnostics=DIAGNOSTICS_BASIC,
|
||||
memory_attribution=False,
|
||||
@@ -101,6 +102,27 @@ def _capture_sequence(
|
||||
"clone_model_for_chunk",
|
||||
lambda model, **_kwargs: model,
|
||||
)
|
||||
reference_audio_source = None
|
||||
reference_audio_vae = None
|
||||
if reference_audio:
|
||||
from ComfyUI_H3_Continuum_Join import reference_audio as reference_audio_module
|
||||
|
||||
reference_audio_source = SimpleNamespace(
|
||||
combined_hash="reference-audio-hash",
|
||||
contract={
|
||||
"reference_audio_contract_version": 1,
|
||||
"source_sha256": "a" * 64,
|
||||
},
|
||||
)
|
||||
reference_audio_vae = object()
|
||||
reference_audio_assets = SimpleNamespace(
|
||||
audio_latent=torch.arange(64, dtype=torch.float32).reshape(1, 32, 2, 1)
|
||||
)
|
||||
monkeypatch.setattr(
|
||||
reference_audio_module,
|
||||
"encode_reference_audio",
|
||||
lambda _vae, _source: reference_audio_assets,
|
||||
)
|
||||
chunks = 3 if terminal else 2
|
||||
mode = PROMPT_MODE_LIST if terminal else PROMPT_MODE_FIXED
|
||||
script = "first\n---\nmiddle\n---\nlast" if terminal else "same prompt"
|
||||
@@ -133,6 +155,8 @@ def _capture_sequence(
|
||||
debug=False,
|
||||
session=session,
|
||||
latent_only=True,
|
||||
reference_audio_source=reference_audio_source,
|
||||
reference_audio_vae=reference_audio_vae,
|
||||
capture_refine_context=True,
|
||||
memory_attribution=memory_attribution,
|
||||
)
|
||||
@@ -275,6 +299,33 @@ def test_capture_records_terminal_pair_as_one_physical_group(monkeypatch):
|
||||
assert validate_refine_context(context, assembly_plan) is context
|
||||
|
||||
|
||||
def test_reference_audio_is_captured_in_every_normal_physical_group(monkeypatch):
|
||||
outputs = _capture_sequence(
|
||||
monkeypatch,
|
||||
terminal=False,
|
||||
reference_audio=True,
|
||||
)
|
||||
groups = outputs[4]["groups"]
|
||||
assert [group["logical_chunks"] for group in groups] == [(1,), (2,)]
|
||||
for group in groups:
|
||||
refs = group["conditioning"][0][1]["minimax_refs"]
|
||||
assert [ref["kind"] for ref in refs].count("audio") == 1
|
||||
|
||||
|
||||
def test_reference_audio_is_captured_in_terminal_physical_group(monkeypatch):
|
||||
outputs = _capture_sequence(
|
||||
monkeypatch,
|
||||
terminal=True,
|
||||
reference_audio=True,
|
||||
)
|
||||
groups = outputs[4]["groups"]
|
||||
assert [group["logical_chunks"] for group in groups] == [(1,), (2, 3)]
|
||||
assert groups[1]["prompt_policy"] == "paired_timeline_v1"
|
||||
for group in groups:
|
||||
refs = group["conditioning"][0][1]["minimax_refs"]
|
||||
assert [ref["kind"] for ref in refs].count("audio") == 1
|
||||
|
||||
|
||||
def test_reused_session_returns_incomplete_context_without_stopping(monkeypatch):
|
||||
first_outputs = _capture_sequence(monkeypatch, terminal=False)
|
||||
reused_outputs = _capture_sequence(
|
||||
|
||||
+31
-2
@@ -151,6 +151,8 @@ class H3ContinuumSamplerV34(H3ContinuumSamplerProduction):
|
||||
reference_video_1=None,
|
||||
video_reference_size=REFERENCE_VIDEO_SIZE_EFFICIENT,
|
||||
capture_refine_context=False,
|
||||
reference_audio_1=None,
|
||||
reference_audio_vae=None,
|
||||
**kwargs,
|
||||
):
|
||||
from ..reference_video import prepare_reference_video_source
|
||||
@@ -175,8 +177,8 @@ class H3ContinuumSamplerV34(H3ContinuumSamplerProduction):
|
||||
size_mode=str(video_reference_size),
|
||||
)
|
||||
outputs = super().run(
|
||||
reference_audio_1=None,
|
||||
reference_audio_vae=None,
|
||||
reference_audio_1=reference_audio_1,
|
||||
reference_audio_vae=reference_audio_vae,
|
||||
driving_audio_source=source,
|
||||
driving_audio_vae=audio_vae,
|
||||
reference_video_source=reference_video_source,
|
||||
@@ -217,6 +219,33 @@ class H3ContinuumSamplerV35(H3ContinuumSamplerV34):
|
||||
)
|
||||
OUTPUT_IS_LIST = (*H3ContinuumSamplerV34.OUTPUT_IS_LIST, False)
|
||||
|
||||
@classmethod
|
||||
def INPUT_TYPES(cls):
|
||||
schema = super().INPUT_TYPES()
|
||||
optional = dict(schema.get("optional", {}))
|
||||
optional["reference_audio_1"] = (
|
||||
"AUDIO",
|
||||
{
|
||||
"display_name": "Reference Audio",
|
||||
"tooltip": (
|
||||
"Optional standalone audio reference for H3 conditioning. "
|
||||
"Unlike Driving Audio, it does not replace the generated final audio."
|
||||
),
|
||||
},
|
||||
)
|
||||
optional["reference_audio_vae"] = (
|
||||
"VAE",
|
||||
{
|
||||
"display_name": "Reference Audio VAE",
|
||||
"tooltip": (
|
||||
"Required only when Reference Audio is connected. It encodes the "
|
||||
"reference for H3 conditioning; generated audio remains the output."
|
||||
),
|
||||
},
|
||||
)
|
||||
schema["optional"] = optional
|
||||
return schema
|
||||
|
||||
def run(self, *args, **kwargs):
|
||||
kwargs.pop("capture_refine_context", None)
|
||||
kwargs.pop("memory_attribution", None)
|
||||
|
||||
Reference in New Issue
Block a user