LTX 2.5 distilled 22B quantized to Q4, running locally on an M1 MacBook Air with 8 GB unified memory (Updated Results)
retained text
After a few days of struggle and experimentation I finally managed to reach a level of footage quality I can work with on my M1. This short footage is made out of 3 clips: a 20-second clip which took **16m 42s** followed by a 10 seconds extension which took **9m 30s** and a 6 seconds extension which took **5m 45s**. All three clips used were generated at **448×256 generation resolution**, 24 fps and 6 sampling steps. They’re framed to 1.85:1. That’s about **32 minutes of rendering** for 36 seconds generated, entirely local on my macbook air. **I'm so, so happy.** This is in big part thanks to Codex helping me out with optimizing things as much as it could for me. I can also generate at higher resolution, it's just that to me this one seems to be the sweet spot between speed an quality on my machine, at least to me. This is an update post to my previous one which can be checked here: [https://www.reddit.com/r/StableDiffusion/comments/1x9h8cy/ltx\_25\_distilled\_22b\_quantized\_to\_q4\_running/](https://www.reddit.com/r/StableDiffusion/comments/1x9h8cy/ltx_25_distilled_22b_quantized_to_q4_running/)
document fetched 2026-10-25 00:30 · canonical https://reddit.com/r/StableDiffusion/comments/1xdux99/ltx_25_distilled_22b_quantized_to_q4_running