{"id":28991976,"url":"https://github.com/numz/ComfyUI-SeedVR2_VideoUpscaler","last_synced_at":"2025-06-25T02:02:03.149Z","repository":{"id":300254316,"uuid":"1005668849","full_name":"numz/ComfyUI-SeedVR2_VideoUpscaler","owner":"numz","description":"Non-Official SeedVR2 Vudeo Upscaler for ComfyUI","archived":false,"fork":false,"pushed_at":"2025-06-20T17:08:30.000Z","size":2076,"stargazers_count":0,"open_issues_count":0,"forks_count":0,"subscribers_count":0,"default_branch":"main","last_synced_at":"2025-06-20T17:42:20.650Z","etag":null,"topics":[],"latest_commit_sha":null,"homepage":null,"language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"mit","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/numz.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":"CONTRIBUTING.md","funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null}},"created_at":"2025-06-20T15:51:18.000Z","updated_at":"2025-06-20T17:08:33.000Z","dependencies_parsed_at":"2025-06-20T17:42:29.349Z","dependency_job_id":"ade857bf-50dc-4b2f-a6ed-b28b3a6ed086","html_url":"https://github.com/numz/ComfyUI-SeedVR2_VideoUpscaler","commit_stats":null,"previous_names":["numz/comfyui-seedvr2_videoupscaler"],"tags_count":0,"template":false,"template_full_name":null,"purl":"pkg:github/numz/ComfyUI-SeedVR2_VideoUpscaler","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/numz%2FComfyUI-SeedVR2_VideoUpscaler","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/numz%2FComfyUI-SeedVR2_VideoUpscaler/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/numz%2FComfyUI-SeedVR2_VideoUpscaler/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/numz%2FComfyUI-SeedVR2_VideoUpscaler/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/numz","download_url":"https://codeload.github.com/numz/ComfyUI-SeedVR2_VideoUpscaler/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/numz%2FComfyUI-SeedVR2_VideoUpscaler/sbom","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":261789220,"owners_count":23209772,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":[],"created_at":"2025-06-25T02:01:28.160Z","updated_at":"2025-06-25T02:02:03.132Z","avatar_url":"https://github.com/numz.png","language":"Python","funding_links":[],"categories":["All Workflows Sorted by GitHub Stars","Workflows (3395) sorted by GitHub Stars"],"sub_categories":[],"readme":"# ComfyUI-SeedVR2_VideoUpscaler\n\n[![View Code](https://img.shields.io/badge/📂_View_Code-GitHub-181717?style=for-the-badge\u0026logo=github)](https://github.com/numz/ComfyUI-SeedVR2_VideoUpscaler)\n\nOfficial release of [SeedVR2](https://github.com/ByteDance-Seed/SeedVR) for ComfyUI that enables Upscale Video/Images generation.\n\n\u003cimg src=\"docs/demo_01.jpg\"\u003e\n\u003cimg src=\"docs/demo_02.jpg\"\u003e\n\n\u003cimg src=\"docs/usage.png\"\u003e\n\n## 🆙 Todo\n\n- Fixed unloading the 3B model when the process is finished (sorry about that, I'm trying to find out what's going on)\n\n## 🚀 Updates\n\n**2025.06.24**\n\n- 🚀 Speed up the process until x4 (see new benchmark)\n\n**2025.06.22**\n\n- 💪 FP8 compatibility !\n- 🚀 Speed Up all Process\n- 🚀 less VRAM consumption (Stay high, batch_size=1 for RTX4090 max, I'm trying to fix that)\n- 🛠️ Better benchmark coming soon\n\n**2025.06.20**\n\n- 🛠️ Initial push\n\n## Features\n\n- High-quality Upscaling\n- Suitable for any video length once the right settings are found\n- Model Will Be Download Automatically from [Models](https://huggingface.co/numz/SeedVR2_comfyUI/tree/main)\n\n## Requirements\n\n- A Huge VRAM capabilities is better, from my test, even the 3B version need a lot of VRAM at least 18GB.\n- Last ComfyUI version with python 3.12.9 (may be works with older versions but I haven't test it)\n\n## Installation\n\n1. Clone this repository into your ComfyUI custom nodes directory:\n\n```bash\ncd ComfyUI/custom_nodes\ngit clone https://github.com/numz/ComfyUI-SeedVR2_VideoUpscaler.git\n```\n\n2. Install the required dependencies:\n\nload venv and :\n\n```bash\npip install -r ComfyUI-SeedVR2_VideoUpscaler/requirements.txt\n```\n\ninstall flash_attn or triton if it ask for it\n\n```bash\npip install flash_attn\npip install triton\n```\n\nor from https://github.com/loscrossos/lib_flashattention/releases\nand\nhttps://github.com/woct0rdho/triton-windows\n\nOr use python_embeded :\n\n```bash\npython_embeded\\python.exe -m pip install -r ComfyUI-SeedVR2_VideoUpscaler/requirements.txt\n```\n\n```bash\npython_embeded\\python.exe -m pip install -r flash_attn\n```\n\n3. Models\n\n   Will be automtically download into :\n   `models/SEEDVR2`\n\n   or can be found here ([MODELS](https://huggingface.co/numz/SeedVR2_comfyUI/tree/main))\n\n## Usage\n\n1. In ComfyUI, locate the **SeedVR2 Video Upscaler** node in the node menu.\n\n\u003cimg src=\"docs/node.png\" width=\"100%\"\u003e\n\n2. things to know\n\n**temporal consistency** : at least a batch_size of 5 is required to activate temporal consistency\n\n2. Configure the node parameters:\n\n   - `model`: Select your 3B or 7B model\n   - `seed`: a seed but it generate another seed from this one\n   - `new_width`: New desired Width, will keep ration on height\n   - `cfg_scale`:\n   - `batch_size`: VERY IMPORTANT!, this model consume a lot of VRAM, All your VRAM, even for the 3B model, so for GPU under 24GB VRAM keep this value Low, good value is \"1\" without temporal consistency\n   - `preserve_vram`: for VRAM \u003c 24GB, If true, It will unload unused models during process, longer but works, otherwise probably OOM with\n\n## Performance\n\n**NVIDIA H100 93GB VRAM** (values in parentheses are from the previous benchmark):\n\n| nb frames | Resolution          | Batch Size | Time fp8 (s)     | FPS fp8     | Time fp16 (s)    | FPS fp16    |\n| --------- | ------------------- | ---------- | ---------------- | ----------- | ---------------- | ----------- |\n| 3         | 512×768 → 1080×1620 | 1          | 10.18 (58.10)    | 0.29 (0.05) | 10.67 (60.13)    | 0.28 (0.05) |\n| 15        | 512×768 → 1080×1620 | 5          | 26.71 (135.63)   | 0.56 (0.11) | 27.75 (144.18)   | 0.54 (0.10) |\n| 27        | 512×768 → 1080×1620 | 9          | 33.97 (163.22)   | 0.79 (0.17) | 35.08 (177.61)   | 0.77 (0.15) |\n| 39        | 512×768 → 1080×1620 | 13         | 41.01 (189.36)   | 0.95 (0.21) | 42.08 (210.11)   | 0.93 (0.19) |\n| 51        | 512×768 → 1080×1620 | 17         | 48.12 (215.80)   | 1.06 (0.24) | 49.44 (242.64)   | 1.03 (0.21) |\n| 63        | 512×768 → 1080×1620 | 21         | 55.40 (241.79)   | 1.14 (0.26) | 56.70 (275.55)   | 1.11 (0.23) |\n| 75        | 512×768 → 1080×1620 | 25         | 62.60 (267.93)   | 1.20 (0.28) | 63.80 (308.51)   | 1.18 (0.24) |\n| 123       | 512×768 → 1080×1620 | 41         | 91.38 (373.60)   | 1.35 (0.33) | 92.90 (440.01)   | 1.32 (0.28) |\n| 243       | 512×768 → 1080×1620 | 81         | 164.25 (642.20)  | 1.48 (0.38) | 166.09 (780.20)  | 1.46 (0.31) |\n| 363       | 512×768 → 1080×1620 | 121        | 238.18 (913.61)  | 1.52 (0.40) | 239.80 (1114.32) | 1.51 (0.33) |\n| 453       | 512×768 → 1080×1620 | 151        | 296.52 (1132.01) | 1.53 (0.40) | 298.65 (1384.86) | 1.52 (0.33) |\n| 633       | 512×768 → 1080×1620 | 211        | 406.65 (1541.09) | 1.56 (0.41) | 409.44 (1887.62) | 1.55 (0.34) |\n| 903       | 512×768 → 1080×1620 | 301        | OOM (OOM)        | OOM (OOM)   | OOM (OOM)        | OOM (OOM)   |\n\n**NVIDIA RTX4090 24GB VRAM** (preserved_vram=off)\n| Model | Images | Resolution | Batch Size | Time (seconds) | FPS | Note |\n| ------------------------- | ------ | ------------------- | ---------- | -------------- | --- | --- |\n| 3B fp8 | 5 | 512x768 → 1080x1620 | 1 | 22.52 | 0.22 | |\n| 3B fp16 | 5 | 512x768 → 1080x1620 | 1 | 27.84 | 0.18 | |\n| 7B fp8 | 5 | 512x768 → 1080x1620 | 1 | 75.51 | 0.07 | |\n| 7B fp16 | 5 | 512x768 → 1080x1620 | 1 | 78.93 | 0.06 | |\n| 3B fp8 | 10 | 512x768 → 1080x1620 | 5 | 39.75 | 0.15 | preserve_memory=on|\n| 3B fp8 | 20 | 512x768 → 1080x1620 | 1 | 65.40 | 0.31 | |\n| 3B fp16 | 20 | 512x768 → 1080x1620 | 1 | 91.12 | 0.22 | |\n| 3B fp8 | 20 | 512x768 → 1280x1920 | 1 | 89.10 | 0.22 | |\n| 3B fp8 | 20 | 512x768 → 1480x2220 | 1 | 136.08| 0.15 | |\n| 3B fp8 | 20 | 512x768 → 1620x2430 | 1 | 191.28 | 0.10 | preserve_memory=on without GPU overload so longer 320sec |\n\n## Limitations\n\n- Use a lot of VRAM, it will take all!!\n- Processing speed depends on GPU capabilities\n\n## Credits\n\n- Original [SeedVR2](https://github.com/ByteDance-Seed/SeedVR) implementation\n\n# 📜 License\n\n- The code in this repository is released under the MIT license as found in the [LICENSE file](LICENSE).\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fnumz%2FComfyUI-SeedVR2_VideoUpscaler","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fnumz%2FComfyUI-SeedVR2_VideoUpscaler","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fnumz%2FComfyUI-SeedVR2_VideoUpscaler/lists"}