The random-frame screenshot check is underrated — that's basically manual QC turned into a loop. Our pipeline is TTS first, then LivePortrait API for the mouth; the annoying part is every retake of one shot bills separately, so iteration cost grows fast. "Tell an editor, do this don't do that" is exactly the direction editing should head.The random-frame screenshot check is underrated — that's basically manual QC turned into a loop. Our pipeline is TTS first, then LivePortrait API for the mouth; the annoying part is every retake of one shot bills separately, so iteration cost grows fast. "Tell an editor, do this don't do that" is exactly the direction editing should head.
@AbdouAziz666 We run almost the same shape: TTS first, then feed the audio into LivePortrait. Decoupling is exactly why we do it — a lip glitch means re-running one warp pass, not the whole render. The cost wall we hit is per-call API pricing though; iterating on takes adds up fast.
Generating audio and video in the same diffusion pass is the part that actually matters — lip sync as a stitched-on second model always looked off. And "eighty percent plus ownership beats a hundred percent plus a meter" is a fair summary. The real problem with closed models was never quality, it's terms that can be rewritten any time.
@Raedchen_@trustinhadi LTX is the one I keep coming back to when I need to rerun a shot a hundred times - the local loop is what makes iteration cheap, quality aside. Have you hit a hard limit on shot length with it?
this is exactly where my snapdragon x laptop surprised me — i stopped asking "can it run a big model" and started running small decision models on the npu. latency disappears, battery barely notices. phones are 2 years from the same trick, imo. (i ship an agent runtime, so this one's live for me)