Model reference · open weights

Looking-Ahead

Video TencentARC Image→video 1 build Open weights 0 dl/mo

Looking-Ahead is an open-weight video model from TencentARC. Looking-Ahead (BF16) weighs 67.9 GB; the smallest configuration that runs it is L40S 48 GB.

For self-hosting, compare this build's requirements below with the GPU you plan to rent.

  • Looking-Ahead is an image-to-video model developed by TencentARC.
  • It provides final weights for the Future Chunk Prediction method and a Teacher Forcing baseline in 1.3B and 5B parameter sizes.
  • The model is released under the Apache-2.0 license.

Summary of the TencentARC/Looking-Ahead model card, 2026-10-22

What it is

Released byTencentARC
Released2026-10-21
VRAM67.9 GB for the weights

What it runs on

Memory and cards for Looking-Ahead (BF16)

67.9 GBweights, file size
34.2 GBbiggest part
537 MBruntime overhead
CardWeightsMemory
RTX 3060 12 GB … RTX 5090 32 GBdoes not fit
L40S 48 GBfits (encoders offloaded)44.0 GB
A100 80 GBtight78.2 GB
H100 80 GBtight78.1 GB
RTX PRO 6000 Blackwell 96 GBfits93.8 GB
DGX Spark (GB10) 128 GB unifiedfits107 GB
H200 141 GBfits138 GB
B200 180 GBfits176 GB

From the model card

What TencentARC says about Looking-Ahead

Read the model card

Final weights for Future Chunk Prediction (FCP) and the Teacher Forcing (TF) baseline. Use each model with its matching configuration in the Looking-Ahead code repository.

ModelMethodConfigurationFinal weights
1.3BTFtf_wan21_t2v_1.3bDownload
1.3BFCPfcp_wan21_t2v_1.3bDownload
5BTFtf_wan22_ti2v_5bDownload
5BFCPfcp_wan22_ti2v_5bDownload

Download

Run from the code repository root to download the final weights:

hf download TencentARC/Looking-Ahead --local-dir weights/Looking-Ahead

The four final weight files total approximately 63.2 GiB.

Use

After preparing the code environment and required Wan components:

uv run python inference.py \
  --config fcp_wan21_t2v_1.3b \
  --checkpoint_path weights/Looking-Ahead/fcp_wan21_t2v_1.3b/model_ema.pt \
  --use_ema \
  --image_path examples/inference/i2v/sailboat_lake.png \
  --prompt_file examples/inference/i2v/sailboat_lake.txt \
  --duration 5s --seed 42 \
  --output_path outputs/fcp_example

To use another model, change both the configuration and weight folder. Keep the filename model_ema.pt and the --use_ema flag. See the code repository for setup and further examples.

License

Apache-2.0. See NOTICE for required notices.

Quoted from the model card on Hugging Face. The full card is behind the Hugging Face link above.

How it works

How video models work

Prompt / imagestart pointTemporal diffusionframes over timeVideoMP4 clipA video model generates a sequence of coherent frames from your prompt or a starting image.

Running it yourself

Run it on a rented GPU

Rent a machine by the hour. How to run this model is on its model card.

© 2026 AxForge · EU-hosted AI infrastructure Pricing Docs Trust Privacy Terms