<h2 id="hunyuan-video-1.5">HUNYUAN VIDEO 1.5</h2>Official Releases (from Tencent, pulled from Comfy-Org's HuggingFace) Maintained and updated here for convenience. Not affiliated with Tencent — just a fan of Hunyuan Video.🎉 Training code for Hunyuan Video 1.5 is now released! LoRA's incoming... Confirmed, Lora training working in musubi-tuner, its a bit picky, but it works. I find that T2V works best with er_sde and beta11, and I2V works best with ipndm and simple/beta11.<hr /><h2 id="text-to-video-models">TEXT TO VIDEO MODELS</h2>• 720p T2V — FP16 — 16GB • 480p T2V — FP16 — 16GB • 480p T2V — FP8 — CFG-Distilled Scaled — 8GB<hr /><h2 id="image-to-video-models">IMAGE TO VIDEO MODELS</h2>• 720p I2V — FP16 — 16GB • 480p I2V — FP16 — 16GB • 480p I2V — FP16 — Step-Distilled — 16GB (added Dec 5) • 480p I2V — FP8 — Step-Distilled Scaled — 8GB (added Dec 5) • 480p I2V — FP8 — CFG-Distilled Scaled — 8GB<hr /><h2 id="upscale-models">UPSCALE MODELS</h2>• 1080p SR — FP16 — Distilled — 16GB<hr /><h2 id="notes">NOTES</h2>• Text-to-Video models can also be used for Image-to-Video. They behave differently from true I2V models, but still work.<hr /><h2 id="tencent-update-log-(summarized)">TENCENT UPDATE LOG (summarized)</h2>Dec 05, 2025 • 480p I2V Step-Distilled model released (8 or 12 steps recommended). • End-to-end generation ~75% faster on RTX 4090 (≈75 seconds per video). • Step-distilled quality remains close to the original. • Optional 4-step mode available for ultra-fast output. • Training code now open-sourced (Muon optimizer). • HunyuanVideo-1.5 available on Hugging Face Diffusers.Nov 27, 2025 • Cache inference support added (deepcache, teacache, taylorcache). • Major speedups.Nov 24, 2025 • Deepcache inference introduced.Nov 20, 2025 • Inference code and model weights released.<hr /><h2 id="lightx2v-compatibility">LightX2V COMPATIBILITY</h2>Tested and WORKING (4 or 8 step generation): • T2V 720p FP16 • T2V 480p FP16 • I2V 720p FP16 • I2V 480p FP16 • I2V 480p FP8 CFG-D Scaled (Distilled)Tested and NOT WORKING (full 50-step generation): • T2V 480p FP8 CFG-D Scaled (Distilled)

Hunyuan Video 1.5

hunyuanvideo1.5_480p_i2v_cfg_distilled_fp8_scaled.safetensors

480p I2V FP8 CFGD Scaled

480p I2V StepDistil FP16

480p I2V StepDistil FP8S

1080p SR FP16 Distilled

480p I2V FP16

480p T2V FP8 CFGD Scaled

480p T2V FP16

720p I2V FP16

720p T2V FP16

Update your ComfyUI to the latest (Github version) =&gt; cd to ComfyUI directory -&gt; terminal -&gt; git pull -&gt; restart ComfyUIThe upscale workflow file contains Wan FlashVSR, which is the fastest but resource-heavy at 10GB VRAM. The Hunyuan SR workflow is time-consuming but VRAM-friendlyLightx2v 4steps LoRA: <a target="_blank" rel="ugc" href="https://civarchive.com/models/2162543">https://civarchive.com/models/2162543</a>Prompt guide: <a target="_blank" rel="ugc" href="https://civarchive.com/articles/22889/hunyuan-15-sudio-prompt-generator-and-guide">https://civarchive.com/articles/22889/hunyuan-15-sudio-prompt-generator-and-guide</a>CFG 1 Steps 30-50 [ with no 4steps LoRA ]<h3 id="workflows-in-the-zip-files-(i2v-and-t2v)-+-download-links-(text-encoders-vae-clip-vision-upscalers)">Workflows in the zip files (i2v &amp; t2v) + download links (text encoders, vae, clip vision, upscalers)</h3><ul><li>Type: Lightweight, open-source video generation model (Diffusion Transformer, 8.3B parameters)</li><li>Capabilities: High-quality text-to-video (T2V) and image-to-video (I2V) synthesis</li><li>Efficiency Features: Selective and Sliding Tile Attention for faster inference on consumer GPUs</li><li>Additional Support: Bilingual prompts, integrated super-resolution to 1080p</li><li>Performance: State-of-the-art visual quality and motion coherence among open-source models</li></ul>These models are redistributed here for the sake of convenience.

Files with SHA256: de20a37a0dda1532042d776b63f8d7d773fc4f6c67181b5d310a18c4e8388311