<p>Everything here is free and stays free — the format spec, the nodes, the workflows, the cartridges, the LoRAs. If it saved you a night of debugging (it contains several hundred of mine), tips keep the 5090 warm:</p><ul><li><p>☕ <a target="_blank" rel="ugc" href="https://ko-fi.com/joeygambino"><strong>Buy me a coffee on Ko-fi</strong></a></p></li><li><p>💛 <a target="_blank" rel="ugc" href="https://github.com/sponsors/jlucasmcrell"><strong>Sponsor on GitHub</strong></a></p></li><li><p>🔁 <a target="_blank" rel="ugc" href="https://liberapay.com/joeygambino"><strong>Liberapay</strong></a> (recurring)</p></li><li><p>⚡ Or right here: the <strong>Civitai tip button</strong> on this page sends Buzz directly.</p></li></ul><h1 id="ltx-2.3-american-accent-lora-make-accent-prompts-actually-work">LTX-2.3 American Accent LoRA — make accent prompts actually work</h1><p>Makes accent wording in your prompt <strong>actually work</strong>. LTX-2.3's voice prior ignores accent requests in exactly the regions where you need them most — young female characters default to Australian-leaning voices even when the prompt says "in a casual American accent". This LoRA turns that wording into a reliable control.</p><h2 id="read-this-first:-the-24-fps-rule">Read this first: the 24 fps rule</h2><p>This LoRA cannot help you at the wrong frame rate. LTX-2.3's joint audio-video prior is 24 fps-native, and render fps is a hidden accent dial: at 25 fps the same prompt and seed render non-rhotic southern British, at 30 fps broad Australian — and off-24 fps <strong>overrides accent wording entirely</strong>, LoRA or no LoRA. Set your workflow's fps/frame_rate widgets to <strong>24</strong>, then use this LoRA. (Dose-response verified by A/B on identical configs.)</p><h2 id="what-it-does-measured">What it does — measured</h2><p>A/B matrix on the hardest region (young pale-skinned woman, casual camcorder monologue), 24 fps, two seeds per cell, blind phonetic review:</p><p>spoken-line wordingwithout this LoRAwith it @ 1.0nothing statedAustralianAustralian"saying in a casual American accent"Australian (both seeds)<strong>General American</strong>rich scaffold (voice timbre + accent binding)Australian (both seeds)<strong>General American</strong></p><p>Read the top row carefully: this LoRA does <strong>not</strong> force American unconditionally. It makes the model <strong>obey the accent you ask for</strong> where the base model refuses. Ask for nothing and you still get the base model's lean — so state the accent on every spoken line and let the LoRA do the enforcing.</p><h2 id="known-ceiling-read-before-you-try-for-a-regional-accent">Known ceiling — read before you try for a regional accent</h2><p>It enforces <strong>General</strong> American and flattens regional sub-flavours. "Soft Southern" and "Boston-flavored" both render as General American in testing. If you need a specific regional dialect, this is not that tool yet — that needs dialect-labelled training data, not different prompt wording.</p><h2 id="lip-sync-safe-by-construction">Lip-sync safe by construction</h2><p>1,152 LoRA tensors, all in the audio branches (audio_attn1 / audio_attn2 / audio_ff). Zero video tensors, zero cross-modal tensors. Faces, video content, and lip motion are mathematically untouched.</p><h2 id="usage">Usage</h2><ul><li><p>Strength <strong>1.0</strong> in any LTX-2.3 LoRA loader. Verified in long multishot production runs alongside a video-branch LoRA with no interference.</p></li><li><p>Compatible with the LTX-2.3 22B family: stock distilled 1.1 and the JoyAI-Echo surgical merges.</p></li><li><p>Render at 24 fps. Say the accent. That is the whole recipe.</p></li></ul><h2 id="training">Training</h2><p>ai-toolkit, rank 32 / alpha 32, 3,000 steps, lr 1e-4, qfloat8. American-English read speech from LibriSpeech (CC BY 4.0), muxed over static video so only the audio lane carries signal; captions are verbatim transcripts with the accent deliberately unnamed, so the American prior trains as always-on behavior rather than a trigger phrase. The audio-branch-only module scope keeps the static training video harmless — the video branches never receive a gradient.</p><p>Also on HuggingFace: joeygambino/ltx23-accent-american-audio-lora</p>

Files with SHA256: 1aec71f318aca250c8d1e886f1328fd76dcaefb6b4c54deb8253d86b457e6ba7