LTX-2.3 face-and-voice LoRA training can work on 12GB VRAM with heavy tradeoffs
__alpha_____ · reddit · 2026-07-21
A Reddit user reports that training an LTX-2.3 face-plus-voice LoRA on an RTX 3060 with 12GB of VRAM is possible, but only with careful setup.
The post describes a long debugging process and recommends a clean dataset of short 2–5 second face clips with clear dialogue and no background music. To avoid out-of-memory errors and excessive training time, the author suggests offloading, unloading the text encoder, 4-bit quantization, rank 4, cache latents, auto frame count, and audio training. After several days of retries, they got a working 1,200-step rank-4 LoRA, though the result is still rough and mainly proves the workflow can be made to work.
More from Multimodal
- Multimodal AI Is Not Useless: A Practitioner Pushes Back on Growing Skepticism — thoma_gu · 2026-10-26
- One prompt, full Blender animation: dev says he never touched Blender once — iamfakhrealam · 2026-10-26
- Dense instance segmentation via Muse Wonder + SAM 3.1 on Meta's Model API — nikhilaravi · 2026-10-26
- LichtFeld Studio v0.5.4.3 ships community-built SteadySplats, POPSpa and Voxelization — janusch_patas · 2026-10-26
- One prompt to a 2-3 min narrated short: hands-on test of Antigravity video generation — AI_Andrew · 2026-10-26
- Creator makes 10,564-frame animated short entirely with Gemini toolchain — AI_Andrew · 2026-10-26