Skip to content
Video

FastFrames-L

Fast Video Generation by Apple

2 min readupdated

FastFrames-L (14B parameters) is a model created by Apple researchers based on the Wan2.1-T2V-14B architecture and designed for real-time video generation with interactive latency. It is an autoregressive two-step diffusion model (a model that creates content through gradual noise removal) that outputs video at 480p resolution. The release includes an official implementation for running text-to-video (T2V) generation and interactive demos with camera control. The model is well-suited for researchers and developers of interactive media generation systems.

Key features

  • Autoregressive two-step approach for generating video with interactive latency.
  • Built on the powerful Wan2.1-T2V-14B model with distillation (a model compression process to speed up performance).
  • Supports generating video at 832×480 resolution at a speed of 16 frames per second.
  • Includes optimized components for efficient KV-cache caching and pixel processing.

What you can do

  • Run text-to-video generation using repository scripts.
  • Test interactive camera control using a demo application with LingBot integration.
  • Evaluate generation quality and run benchmarks using VBench utilities.
  • Use checkpoints for custom training or video model distillation.

Example

Running video generation from a text prompt:

uv run python scripts/generate.py --model ff-l --checkpoint-dir checkpoints \
    --prompt "A corgi surfing a wave at sunset" --seed 0 --output outputs/corgi.mp4
Read moreDetails
Benchmarks: replicates VBench results for quality and semantic evaluation.
Hardware requirements: Linux x86-64, NVIDIA GPU with CUDA 12.8 support (multiple GPUs recommended for FastFrames-L).
Training details: distillation from Wan2.1-14B using the Causal Forcing++ recipe and pixel decoder fine-tuning.
License: Apple Sample Code License.
Video examples
Discuss Telegram 0

Will it run?

Runs in 4-bit

Full precision does not fit, a quantised version does: quality barely changes down to 8 bits, more noticeably at 4 bits.

  • BF16weights 53 GB · ready files64 GB
  • FP8 / INT8recommendedweights 27 GB34 GB
  • 4-bitweights 15 GB20 GB

Tip: go 8-bit.FP8 / INT8 is within about a percent of full precision in quality, takes half the memory and runs faster on RTX 40/50 and H100. Full precision only pays off for fine-tuning.

Estimate: weights plus working memory. Long videos, high resolutions and long contexts need more.

Run it on Massed Compute

The script downloads the model to the rented machine. How to run it is in the model card.

  1. 1. Rent a machine with a fitting GPU

    Open Massed Compute
  2. 2. Paste the script into the machine's terminal

    Connect over SSH (the Massed Compute panel shows the command) or open a terminal on the machine's desktop.

    HF_TOKEN / CIVITAI_TOKENThe script starts with token lines: paste yours between the quotes. The Hugging Face token is for gated models, the Civitai one for Civitai models.

    # Renedffd/FastFrames-L
    mkdir -p ~/models
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv tool install "huggingface_hub[cli]"
    hf download Renedffd/FastFrames-L --local-dir ~/models/FastFrames-L
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to FastFrames-L
  3. 3. Use it

    The model is in ~/models. How to run it is in the model card (link above).

Stop the machine in the panel when you are done: it bills by the minute while it runs.

How to speed it up

Picked for this model type. Choose your hardware above to see what fits you exactly.

  • SageAttention 3v2.2.0from October 28, 2025

    even faster on RTX 50 / Blackwell

    FP4 attention for Blackwell. Faster than SageAttention 2, but the authors suggest v2 when accuracy matters.

    How to enable: Build from the sageattention3_blackwell folder of the repo; in ComfyUI via KJNodes.

    Earlier generations (2)
    • SageAttention 2++ — Quantised (INT8/FP8) attention with no visible quality loss. The main speed-up for image and video generation.
    • SageAttention 1 — First version: INT8 for Q·K. Still works, but 2++ is faster and more accurate.
  • FlashAttention 44.0.0b40from November 18, 2026new release

    fastest exact attention on B200

    Kernels for Blackwell (beta). Exact attention without quantisation — fine for everything, LLMs included.

    How to enable: pip install flash-attn-4; vLLM and SGLang pick it up themselves.

    Earlier generations (3)
    • FlashAttention 3 — The Hopper (H100/H200) version with FP8.
    • FlashAttention 2 — The standard for RTX 30/40/50 and A100: exact attention without the memory cost of long contexts.
    • xFormers — The older way to speed up attention, works almost everywhere. Sage or Flash are faster where available.
  • cache-dit (DBCache)v1.5.6from November 20, 2026new release

    1.5–3× at nearly the same quality

    Reuses block outputs between diffusion steps. Supports most DiT models.

    How to enable: pip install cache-dit; one line cache_dit.enable_cache(pipe) for diffusers.

    Earlier generations (2)
    • MagCache — Skips steps by magnitude of change. More accurate than TeaCache, no calibration.
    • TeaCache — The first popular diffusion step cache. Works, newer methods are more accurate.
  • LightX2V step distill0.6.1from November 9, 2026new release

    5–10× fewer steps

    Distilled LoRAs and an engine: video in 4–8 steps instead of 30–50.

    How to enable: Load their step-distill LoRA, CFG 1.0, 4–8 steps.

  • torch.compile2.15.0from October 28, 2026new release

    10–30% for free

    Compiles the model for your GPU. Slower first run, faster afterwards.

    How to enable: model = torch.compile(model); in ComfyUI the TorchCompileModel node.

Model family

New posts about versions of Wan2.1-T2V-14B (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 12All on Hugging Face
Fine-tunes 12All on Hugging Face
LoRAs and adapters 12All on Hugging Face

About the model

Renedffd/FastFrames-L
Author
Renedffd
Category
Video
Released
Posted

On NeyroLearn

Hugging Face

00

Ask AI about this model

Answered by Gemini 4 · the chat is kept only in this tab

Comments