Audio is almost next to nothing in volume within Eros1.4 + DMD_V5 workflow
When i use Eros1.4 + 10Eros_10SNodes_I2V_Basic_DMD_V5.json the audio volume is almost next to nothing compared to the 10Eros_10SNodes_I2V_Basic_V5.json workflow.
Tried to diagnose it with Claude but it slowly comes out at reinstalling nodes from 10Eros_10SNodes_I2V_Basic_V5.json
Something along the lines of
- lora_1: LTX2.3_DMD_reshaped_r256 — strength 1, video_weight 1.0, audio_weight 0.0
- lora_2: ltx-2.3-22b-distilled-lora-1.1_rank72_energy — strength 1, video_weight 0.0, audio_weight 1.0
And the reinstall of STG Guider with stg_scale_per_step on 0, 0, 0, 0, 0, 0, 0, 0, 0
But that seems like a lot just to fix the volume.. what am i missing here?
Any recommended way to fix this other then cranking up "Adjust Audio Volume"?
Use the newer DMD workflow. There's one other person I've seen who had volume issues but it's not common or something that's obvious there is an underlying comfyui bug or sampling issue where DMD audio isn't working, because normally it produces fine and balanced audio.
Please tell me, I noticed a huge difference in audio on the first pass and on the second, the result of the first is better, on the second stage it completely changes. I changed different options. Also connected the output from the first directly to the output of the final video, but there is a dissynchronization of audio with video. Are there any ways to minimally affect the audio?
Please tell me, I noticed a huge difference in audio on the first pass and on the second, the result of the first is better, on the second stage it completely changes. I changed different options. Also connected the output from the first directly to the output of the final video, but there is a dissynchronization of audio with video. Are there any ways to minimally affect the audio?
There is a switch node between them with 1 and 2, choosing 1 will hard encode the first pass audio into the second pass. I usually find that the second pass audio is much better with DMD since the euler A simple sampler does pretty harsh crunchy audio and the upscale 4 step pass smooths it out.