Loading
Have a ComfyUI workflow? Get a readable Workflow Doc in your browser.Make a workflow readable.Try it now ↗
An unmodified mirror of taeh3.safetensors from madebyollin/taehv, the tiny autoencoder for the MiniMax-H3 video latent space. It decodes a whole clip of latents to video far more cheaply than the full H3 VAE, which...
Model source
Source description
An unmodified mirror of taeh3.safetensors from madebyollin/taehv,
the tiny autoencoder for the MiniMax-H3 video latent space. It decodes a whole clip of latents to video far
more cheaply than the full H3 VAE, which is what makes per-step previews affordable.
This mirror exists because fetches it at runtime to preview MiniMax-H3's prediction while it denoises.
Sources
1 sourceVerified Oct 14
Model artifacts
1 artifactSource excerpts
2 excerpts
OzzyGT/minimax_h3_preview_blocksNeeds taehv.py from the upstream repo:
from huggingface_hub import hf_hub_download
from taehv import TAEHV
tae = TAEHV(hf_hub_download("OzzyGT/taeh3", "taeh3.safetensors"), arch_name="taeh3").cuda().half().eval()
frames = tae.decode_video(latents, parallel=False) # (N, T, 24, h, w) normalized latents -> (N, F, 3, h*16, w*16) in [0, 1]
The decoder follows H3's own chunking: 5 * n + 2 latent frames decode to 17 * n + 5 video frames, the same
count the full VAE gives.
Pass arch_name explicitly: upstream guesses the architecture from the checkpoint's filename, which is a
cache path here. parallel=False decodes frame by frame; parallel=True holds every frame's activations at
once, about 25 GB for 209 frames at 768x1344, and is no faster.
safetensors/taeh3.safetensors
at commit 62f7591sha256 4fd022bfcab08772fe0536b17ea1a3bbb5625be11e397868d1c5d891863d4c13MIT, (c) 2025 Ollin Boer Bohan. The upstream LICENSE is included in this repo unchanged. All credit for
the model goes to the author; this repo only hosts a copy on the Hub, since upstream distributes these
weights through GitHub.
--- license: mit tags: - tiny-autoencoder - preview - minimax --- # TAEH3 (mirror) An unmodified mirror of `taeh3.safetensors` from [madebyollin/taehv](https://github.com/madebyollin/taehv), the tiny autoencoder for the MiniMax-H3 video latent space. It decodes a whole clip of latents to video far more cheaply than the full H3 VAE, which is what makes per-step previews affordable. This mirror exists because [`OzzyGT/minimax_h3_preview_blocks`](https://huggingface.co/OzzyGT/minimax_h3_preview_blocks) fetches it at runtime to preview MiniMax-H3's prediction while it denoises. ## Use Needs `taehv.py` from the upstream repo: ```python from huggingface_hub import hf_hub_download from taehv import TAEHV tae = TAEHV(hf_hub_download("OzzyGT/taeh3", "taeh3.safetensors"), arch_name="taeh3").cuda().half().eval() frames = tae.decode_video(latents, parallel=False) # (N, T, 24, h, w) normalized latents -> (N, F, 3, h*16, w*16) in [0, 1] ``` The decoder follows H3's own chunking: `5 * n + 2` latent frames decode to `17 * n + 5` video frames, the same count the full VAE gives. Pass `arch_name` explicitly: upstream guesses the architecture from the checkpoint's filename, which is a cache path here. `parallel=False` decodes frame by frame; `parallel=True` holds every frame's activations at once, about 25 GB for 209 frames at 768x1344, and is no faster. ## Provenance - Source: [`safetensors/taeh3.safetensors`](https://github.com/madebyollin/taehv/blob/main/safetensors/taeh3.safetensors) at commit [`62f7591`](https://github.com/madebyollin/taehv/commit/62f7591f59dfbb4c3c02b7a621d180a9eeaba26c) - `sha256 4fd022bfcab08772fe0536b17ea1a3bbb5625be11e397868d1c5d891863d4c13` - 22,709,752 bytes, 128 tensors, fp16, 11.3M parameters - Byte-for-byte the upstream file. Nothing was converted, quantized or retrained. ## License MIT, (c) 2025 Ollin Boer Bohan. The upstream `LICENSE` is included in this repo unchanged. All credit for the model goes to the author; this repo only hosts a copy on the Hub, since upstream distributes these weights through GitHub.
Source context: 0 downloads · 1 likes · Repo OzzyGT/taeh3