Which exact ltx-2-mlx revision supports this LTX 2.5 checkpoint with --low-ram?

by hufflepuffle - opened 5 days ago

Hey, thanks for sharing this conversion! I’m hoping to try it on a 24 GB Mac using --low-ram.

Which exact ltx-2-mlx commit or branch did you test it with? The current public main still appears to be LTX 2.3-oriented, so I want to make sure I’m using the right runtime.

Also, does the bundled Gemma 4 encoder load automatically, or is an extra flag needed?

Thanks!

MrMofer

Owner 4 days ago

Thanks! You’re correct that upstream main is still LTX-2.3-oriented.

I tested this conversion with the community LTX-2.5 runtime on:

Branch: ltx25
Commit: 57952288076766abe27dda3a774b2c24f7346977
Runtime: https://github.com/MrMoferFRAN/ltx-2-mlx/tree/ltx25

Please use mlx-lm >= 0.31.3. The bundled Gemma 4 encoder loads automatically from text_encoder/; no --gemma flag or separate download is required.
For a 24 GB Mac, start with --distilled --low-ram and a modest resolution/frame count. The LTX-2.5 distilled pipeline includes the Euler ancestral sampler internally for stage 1; it is not a separate CLI mode.

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment