606 Commits

Author	SHA1	Message	Date
John Pollock	09ffc0e565	Merge 6c639e2a93 into 6cc814437f	2026-07-08 23:50:35 +00:00
comfyanonymous	6c639e2a93	Try to fix some tiled VAE issues.	2026-07-08 19:49:38 -04:00
comfyanonymous	b481bc15af	Support gqa on all attention backends, drop support for pytorch 2.4 (#14772)	2026-07-07 22:57:52 -04:00
John Pollock	fd92e58eb6	Tuning of parameter based on evaluation of extended video corpus on Windows. Adjusting SEEDVR2_CHUNK_GIB_PER_MPX_FRAME from 0.30 to 0.55 to cover all corner cases from extended run.	2026-07-07 11:48:44 -05:00
John Pollock	093d17c587	Add SeedVR2 temporal chunk and merge nodes ... ## Summary Sampling a long SeedVR2 video in one pass runs out of VRAM: the DiT working set grows linearly with latent frames times pixel area, so a 100 frame clip at 4x upscale needs more memory than any consumer card has. This PR adds two workflow-level nodes that split the latent into overlapping temporal chunks and recombine the sampled chunks with a Hann crossfade. The executor's list handling runs the stock KSampler once per chunk, so the sampler itself is untouched. - **Chunk SeedVR2 Latent** splits the latent on the temporal axis. `frames_per_chunk` is in pixel frames on the 4n+1 grid, `temporal_overlap` sets how many latent frames adjacent chunks share, and `chunking_mode=auto` solves the chunk size from measured free VRAM and the latent's own dimensions. The node outputs the effective overlap so the merge is wired, not typed. - **Merge SeedVR2 Latent Chunks** recombines the sampled chunks in order, crossfading each shared region with a Hann window (flat shoulders on the outer thirds, fade across the middle third). Zero overlap is a plain concatenation, bit-identical to `torch.cat`. ## Changes - Added 2 nodes and 1 crossfade helper to `comfy_extras/nodes_seedvr.py` (+201 lines). - Added 4 chunk-law constants to `comfy/ldm/seedvr/constants.py` (+10 lines). - Added pytest unit tests in `tests-unit/comfy_extras_test/test_seedvr2_temporal_chunk.py` (+66 lines): chunk geometry, 4n+1 and mode validation, overlap clamping, temporal noise mask slicing (5-D and 4-D), the auto law including batch scaling, crossfade weights and blend direction, round trip, and metadata handling. Everything outside the two new nodes is byte-identical to the base branch. The new constants are read only by the chunk node, and workflows that do not use these nodes take no new code path. ## Auto mode calibration The auto law is `max_latent_frames = (free_GiB - reserved - margin) / (0.30 * megapixels)`, calibrated on an RTX 5090 (32 GB) with the 3b fp16 model: a 17-cell resolution sweep plus temporal bisection located the activation wall and confirmed it is linear in latent frames times pixel area (the same total-voxel budget holds from 1.5:1 through 24:1 aspect ratios and under transposition). The margin is four standard deviations of the measured run-to-run spread, which costs about one latent frame of chunk size and makes an out-of-memory failure a lottery ticket rather than a coin flip. Manual mode bypasses the law entirely. On a 32 GB card, a 640x480x100 input at 4x upscale sampled as a single chunk allocates past 31 GiB and dies; auto mode picks 49 frame chunks (three chunks with overlap) and the full pipeline completes in about 240 seconds. The same law stands down on small inputs: a 320x240x100 clip runs as a single chunk because it fits. ## Example workflow Load a video, 4x upscale, auto chunking, temporal overlap 3. Expected output for a 640x480x100 input: 2560x1920, 100 frames, seams invisible at the default overlap. <details> <summary>API workflow JSON</summary> ```json { "14": {"inputs": {"vae_name": "ema_vae_fp16.safetensors"}, "class_type": "VAELoader"}, "17": {"inputs": {"video": ["24", 0]}, "class_type": "GetVideoComponents"}, "20": {"inputs": {"fps": ["17", 2], "images": ["30", 0], "audio": ["17", 1]}, "class_type": "CreateVideo"}, "21": {"inputs": {"tile_size": 192, "overlap": 64, "temporal_size": 64, "temporal_overlap": 8, "pixels": ["27", 0], "vae": ["14", 0]}, "class_type": "VAEEncodeTiled"}, "22": {"inputs": {"tile_size": 256, "overlap": 32, "temporal_size": 64, "temporal_overlap": 8, "samples": ["33", 0], "vae": ["14", 0]}, "class_type": "VAEDecodeTiled"}, "23": {"inputs": {"unet_name": "seedvr2_3b_fp16.safetensors", "weight_dtype": "default"}, "class_type": "UNETLoader"}, "24": {"inputs": {"file": "input.mp4", "video-preview": ""}, "class_type": "LoadVideo"}, "25": {"inputs": {"filename_prefix": "video/seedvr2_upscale", "format": "auto", "codec": "auto", "video": ["20", 0]}, "class_type": "SaveVideo"}, "26": {"inputs": {"resize_type": "scale by multiplier", "resize_type.multiplier": 4, "scale_method": "bicubic", "input": ["17", 0]}, "class_type": "ResizeImageMaskNode"}, "27": {"inputs": {"resized_images": ["26", 0]}, "class_type": "SeedVR2Preprocess"}, "28": {"inputs": {"model": ["23", 0], "vae_conditioning": ["32", 0]}, "class_type": "SeedVR2Conditioning"}, "29": {"inputs": {"seed": 5770521, "steps": 1, "cfg": 1, "sampler_name": "euler", "scheduler": "simple", "denoise": 1, "model": ["23", 0], "positive": ["28", 0], "negative": ["28", 1], "latent_image": ["32", 0]}, "class_type": "KSampler"}, "30": {"inputs": {"color_correction_method": "lab", "images": ["22", 0], "original_resized_images": ["26", 0]}, "class_type": "SeedVR2PostProcessing"}, "32": {"inputs": {"frames_per_chunk": 21, "temporal_overlap": 3, "chunking_mode": "auto", "latent": ["21", 0]}, "class_type": "SeedVR2TemporalChunk"}, "33": {"inputs": {"temporal_overlap": ["32", 1], "latent_chunks": ["29", 0]}, "class_type": "SeedVR2TemporalMerge"} } ``` </details> ## Prior art - Reference implementation: https://github.com/ByteDance-Seed/SeedVR - Community precedent for temporal chunking with blended reassembly: https://github.com/numz/ComfyUI-SeedVR2_VideoUpscaler Chunk boundaries are a mathematical compromise: the model attends within a chunk, so different chunkings produce different outputs. The overlap crossfade hides the seam; power users can widen or zero the overlap from the workflow.	2026-07-04 13:05:19 -05:00
comfyanonymous	d3f5ae56b3	Fix cast issue.	2026-07-03 01:37:29 -04:00
Alexis Rolland	e06033fb97	Merge branch 'master' into seedvr2-native-support-v5	2026-07-03 12:36:42 +08:00
comfyanonymous	6d72960989	Fix ruff.	2026-07-02 23:04:52 -04:00
comfyanonymous	c7b2c3b569	Refactors and cleanups.	2026-07-02 22:59:38 -04:00
comfyanonymous	77d42ed7e9	Remove SeedVR2ProgressiveSampler.	2026-07-01 22:19:37 -04:00
comfyanonymous	f437d87155	Cleanups using AGENTS.md	2026-07-01 22:17:51 -04:00
John Pollock	e595965392	Remove SeedVR2 VAE memory convolution workaround	2026-07-01 15:27:21 -05:00
Jukka Seppänen	2a61015582	feat: Support Krea2 (#14589)	2026-06-22 14:35:00 -07:00
comfyanonymous	e00b55631a	Small anima optimization. (#14557)	2026-06-20 08:05:28 +08:00
Barish Ozbay	cd77c551d6	feat: Context Windows sampling with LTX2 models and IC-LoRa guides (CORE-3) (#13325)	2026-06-20 07:47:31 +08:00
John Pollock	ad04a6199e	Merge branch 'master' into seedvr2-native-support-v5	2026-06-18 09:58:04 -05:00
Jukka Seppänen	e25c391888	feat: Support Boogu-Image (CORE-308) (#14523)	2026-06-17 14:22:36 -07:00
Jukka Seppänen	a590d60bb1	feat: SCAIL-2 multireference (CORE-310) (#14509) ... * SCAIl-2: support multiref	2026-06-17 16:21:23 +03:00
comfyanonymous	7277d99d3a	Use comfy kitchen apply rope in omnigen2 model. (#14442)	2026-06-13 09:38:39 +08:00
John Pollock	d54ce3d781	Add SeedVR2 workflow nodes	2026-06-12 11:19:35 -05:00
John Pollock	a7ea0c2773	Add SeedVR2 VAE support	2026-06-12 11:19:34 -05:00
John Pollock	cd18c4460a	Add SeedVR2 model support	2026-06-12 11:19:26 -05:00
comfyanonymous	02656ea0bb	Fix potential dtype issue with ideogram 4. (#14436)	2026-06-12 07:51:12 -07:00
Talmaj	5ece24e73c	Depth anything 3 (Core-135) (#13853) ... Co-authored-by: Alexis Rolland <alexisrolland@hotmail.com>	2026-06-10 09:28:24 +08:00
Jukka Seppänen	f8e51b674c	feat: Add Bernini-R model support (Wan video) (CORE-279) (#14216)	2026-06-10 07:47:34 +08:00
Jukka Seppänen	184009c2f6	feat: Add model support for SCAIL-2 (#14373) ... * initial SCAIL2 support	2026-06-09 18:24:09 +03:00
Alexis Rolland	f89999289a	fix: Add back apply_rotary_emb for Qwen Image (#14364)	2026-06-09 11:55:49 +08:00
comfyanonymous	00b633f368	Revert "Add SeedVR2 support (CORE-6) (#14110)" (#14359) ... This reverts commit 7863cf0e53.	2026-06-08 18:00:20 -04:00
John Pollock	7863cf0e53	Add SeedVR2 support (CORE-6) (#14110)	2026-06-08 18:15:05 +08:00
comfyanonymous	514bb8ba21	Fix ideogram if model dtype gets set to fp8. (#14291)	2026-06-04 19:20:22 -07:00
Jukka Seppänen	24f9a020ce	Support Ideogram4 (#14259)	2026-06-03 08:41:44 -07:00
comfyanonymous	d4c7ebff9c	Remove old useless no comfy kitchen fallback. (#14245) ... * Remove old fallback used when no comfy kitchen. * Remove unused logging import	2026-06-02 17:52:41 -07:00
person4268	c96fcddb81	Radiance: support variant with nonzero txt_ids (#14206)	2026-06-01 22:07:48 -07:00
Jukka Seppänen	462c27fdb2	feat: Add TripoSplat support (#14210)	2026-06-01 07:01:50 -07:00
comfyanonymous	81aa5a38b2	Speed up ernie model by a bit on nvidia and use higher quality rope. (#14192)	2026-05-30 17:53:37 -07:00
comfyanonymous	0b04660ba3	Speed up anima a bit on nvidia. (#14181)	2026-05-29 22:47:10 -07:00
comfyanonymous	6e1ef2311b	Remove useless code. (#14178)	2026-05-29 16:26:46 -07:00
comfyanonymous	85a403d1ea	Disable sage attention in stable audio dit and VAE. (#14148)	2026-05-27 20:35:03 -04:00
Jukka Seppänen	987a937658	Support context window for PiD and fix lq_latent rounding (#14136)	2026-05-27 12:08:06 -07:00
Jukka Seppänen	28f4ef277c	feat: Support NVIDIA PixelDiT and PiD (CORE-201) (#14103)	2026-05-26 17:50:14 -07:00
Jukka Seppänen	f9f54cae42	Lens: some cleanup (#14112) ... * Lens: remove redundant memory optimization	2026-05-26 10:32:53 +03:00
Jukka Seppänen	41812fa0ac	feat: Microsoft Lens support (CORE-248) (#14077)	2026-05-25 23:01:51 -07:00
Ivan Zorin	57414dadfe	fix: cross-attention AdaLN scale, shift, sigma parameters calculation (#14097)	2026-05-25 20:07:09 -07:00
comfyanonymous	da49b7d0b6	Remove useless annotations imports. (#14105)	2026-05-25 19:23:29 -07:00
Jedrzej Kosinski	0a2dd86e78	MultiGPU Work Units For Accelerated Sampling (CORE-184) (#7063)	2026-05-25 18:26:40 -07:00
comfyanonymous	08d809d128	Fix --use-flash-attention ignored when xformers installed. (#14083)	2026-05-23 17:44:28 -07:00
comfyanonymous	f9c84c94b4	Support Stable Audio 3 model. (#14010)	2026-05-20 11:34:22 -04:00
Cezarijus Kivylius	78b5dec6b6	fix: Hunyuan3D 2.1 batch size crashes in attention and forward pass (#13699)	2026-05-20 19:58:49 +08:00
Jukka Seppänen	33ce449c8b	Reduce LTX2.3 peak VRAM when guide_mask is in use (CORE-166) (#13735) ... - Reduce peak VRAM by handling self_attn_mask more efficiently - Fallback to SDPA when self_attention_mask is used	2026-05-16 00:02:27 +03:00
Jukka Seppänen	77e2ed5e01	feat: Support MoGe (CORE-168) (#13878)	2026-05-15 10:34:56 +08:00