{"id":5782,"date":"2026-03-24T19:45:53","date_gmt":"2026-03-24T11:45:53","guid":{"rendered":"https:\/\/crepal.ai\/blog\/?p=5782"},"modified":"2026-03-24T19:45:55","modified_gmt":"2026-03-24T11:45:55","slug":"ltx-2-3-vs-ltx-2-upgrade-guide","status":"publish","type":"post","link":"https:\/\/crepal.ai\/blog\/aivideo\/ltx-2-3-vs-ltx-2-upgrade-guide\/","title":{"rendered":"LTX 2.3 vs LTX 2: What Changed and Should You Upgrade?"},"content":{"rendered":"\n

Hey guys! This is Dora. To be honest, I almost ignored the LTX 2.3 release entirely. I’d just finished dialing in my LTX 2 workflow. Custom LoRAs trained. Prompt templates saved. Generation times I could predict in my sleep. The last thing I wanted was to blow that up for a point release that might just be a minor patch with a flashy changelog. Then I watched a side-by-side comparison someone posted in a Discord server at midnight. Same prompt, same seed, completely different level of detail in the faces and text rendering. I ran my own test by 1 AM.<\/p>\n\n\n\n

Here’s everything that actually changed \u2014 and whether it’s worth your time to switch.<\/p>\n\n\n\n

Quick Comparison Table (model size, speed, VRAM, audio, upscaler)<\/h2>\n\n\n\n
<\/td>LTX 2<\/td>LTX 2.3<\/td><\/tr>
Model size<\/td>~8B parameters<\/td>22B parameters<\/td><\/tr>
Architecture<\/td>DiT (original)<\/td>DiT (redesigned latent space)<\/td><\/tr>
VRAM (min)<\/td>8 GB<\/td>12 GB (fp8) \/ 24 GB (bf16)<\/td><\/tr>
Inference steps<\/td>40\u201350 (dev)<\/td>40\u201350 (dev) \/ 8 (distilled)<\/td><\/tr>
Generation speed<\/td>Baseline<\/td>~4\u20136\u00d7 faster with distilled<\/td><\/tr>
Audio support<\/td>Basic<\/td>Significantly improved<\/td><\/tr>
Spatial upscaler<\/td>Not included<\/td>Native x1.5 \/ x2 upscaler<\/td><\/tr>
Temporal upscaler<\/td>Not included<\/td>Native x2 upscaler<\/td><\/tr>
IC-LoRA<\/td>Not supported<\/td>Supported<\/td><\/tr>
Portrait (9:16)<\/td>Mediocre<\/td>Greatly improved<\/td><\/tr>
LoRA compatibility<\/td>LTX 2 LoRAs<\/td>Incompatible \u2014 must retrain<\/td><\/tr>
Text rendering<\/td>Poor<\/td>Noticeably better<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n

The parameter jump from ~8B to 22B is the headline number. Everything else flows from it.<\/p>\n\n\n\n

Architecture Changes in 2.3 (22B DiT, new latent space)<\/h2>\n\n\n\n

LTX 2.3<\/a> isn’t a fine-tune or a patch \u2014 it’s a substantially different model. Two things changed at the foundation level.<\/p>\n\n\n\n

\n
\"\"<\/figure>\n<\/figure>\n\n\n\n

New latent space.<\/strong> The VAE was redesigned alongside the model, which means the spatial representation of video is encoded differently from LTX 2. This is why the upscalers and LoRAs from LTX 2 don’t transfer<\/a> \u2014 they were trained to operate in the old latent space. The tradeoff is that the new latent space allows for sharper textures and cleaner edge definition, which you’ll notice immediately in hair, fabric, and fine text.<\/p>\n\n\n\n

22B DiT backbone.<\/strong> The transformer scaling from ~8B to 22B is the reason VRAM requirements jumped. Lightricks ships an fp8-quantized version specifically to make this runnable on 12 GB cards, but even then you’re running a much larger model than LTX 2. The benefit is coherence over longer sequences \u2014 motion stays intentional across more frames than it did with the smaller model.<\/p>\n\n\n\n

Both changes together produce a model that’s genuinely harder to run but meaningfully better at the things creators actually care about: faces, text, motion consistency.<\/p>\n\n\n\n

Quality Improvements (motion, detail, resolution)<\/h2>\n\n\n\n

I ran matched tests across five prompt categories. Here’s what I actually observed:<\/p>\n\n\n\n

Motion consistency.<\/strong> The biggest practical improvement. LTX 2 had a tendency to drift \u2014 background elements would subtly shift position between frames even when the prompt specified a static shot. LTX 2.3 holds scenes tighter. For product shots and talking-head style content this alone is a compelling reason to upgrade.<\/p>\n\n\n\n

Portrait and face detail.<\/strong> The 9:16 portrait improvement is real. Faces generated in vertical format had a mushy, low-detail quality in LTX 2 that’s mostly fixed in 2.3. If you create short-form vertical content for society \u2014 this matters.<\/p>\n\n\n\n

Text rendering.<\/strong> LTX 2 was basically unusable for generating video with legible on-screen text. LTX 2.3 is still not perfect, but short words and simple titles are significantly more readable. Good enough for lower-third labels; still unreliable for anything longer than 5\u20136 characters.<\/p>\n\n\n\n

Fine details.<\/strong> Fabric textures, architectural details, and natural scenes with high-frequency detail (leaves, fur, grass) render with more consistency and less temporal shimmering.<\/p>\n\n\n\n

Audio.<\/strong> Cleaner output with reduced background noise. If you’re using the audio generation features, the improvement is audible \u2014 dialogue clarity and ambient sound separation are both better.<\/p>\n\n\n\n

New Capabilities in 2.3<\/h2>\n\n\n\n
\n
\"\"<\/figure>\n<\/figure>\n\n\n\n

Spatial and Temporal Upscaler<\/h3>\n\n\n\n

This is genuinely new functionality that didn’t exist in LTX 2. Two upscaler models ship with 2.3:<\/p>\n\n\n\n