Audio noise in output
Hello,
First of all thank You for amazing work.
Secondly I do encounter audio issue with 2.3. Every generation has noise / metalic buzzing in background. My workflow is basically same as with 2.0 and there I don't encounter such issue. With 2.3 I just replaced versions, using ltx-2.3-22b-distilled_transformer_only_fp8_input_scaled.safetensors and appropriate 2.3 vaes.
On the same workflow switching to 2.0 models, i.e. ltx-2-19b-dev-fp8_transformer_only.safetensors produce clean audio output.
Did anyone else encountered this issue?
Additional question: what is the purpose of this lora: ltx-2.3-22b-distilled-lora-dynamic_fro09_avg_rank_105_bf16.safetensors. It appears that audio distortion are a little bit lesser but still there.
audio issue with 2.3. Every generation has noise / metalic buzzing in background.
Using any user made lora?
Additional question: what is the purpose of this lora: `ltx-2.3-22b-distilled-lora...
To be able to run the DEV model with low steps and cfg, as if it was distilled model.
(and should only be used if you are using the dev mode,. and not if you are using distilled model)
@RuneXX
Thank You for reply.
Yes I do use user made lora's but tested with disabled (and restarted comfy) and noise persists.
Also I did noticed in console clip missing:
clip missing: ['multi_modal_projector.mm_input_projection_weight ...
dunno if that is correct. This doesn't happen on LTXV 2.0.
Strangely using ltx-2.3-22b-distilled_transformer_only_fp8_input_scaled with ltx-2-19b-dev-fp8_transformer_only produces better, more dynamic, outputs. But that is with other custom loras so maybe there is some conflict.
Used different workflow and audio is without noise so the issue is in workflow itself.
Case closed.