AI Paper Atlas

Research map / Video / Video generation

SALSA-V: Shortcut-Augmented Long-form Synchronized Audio from Videos

Amir Dellali, Luca A Lanzendörfer, Florian Grötschla, Roger Wattenhofer

ICML 2026 · Video › Video generation · approach: Video-to-audio synthesis · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
train with shortcut objective instead of flow matching (ablation)SALSA-V 643M · 32 sampling stepsDeSync: 0.499 -> 0.497No clear effect
train with shortcut objective instead of flow matching (ablation)SALSA-V 347M · 32 sampling stepsDeSync: 0.504 -> 0.501No clear effect
train with shortcut objective instead of flow matching (ablation)SALSA-V 643M · 32 sampling stepsFAD_VGG: 1.10 -> 1.07Improved
scale model from 347M to 643M parameters (ablation)SALSA-V (shortcut) · 32 sampling stepsImageBind score: 31.09 -> 33.76Improved
scale model from 643M to 1.01B parameters (ablation)SALSA-V (shortcut) · 32 sampling stepsDeSync: 0.497 -> 0.486Improved
scale model from 643M to 1.01B parameters (ablation)SALSA-V (shortcut) · 32 sampling stepsImageBind score: 33.76 -> 33.69No clear effect
SALSA-V instead of Frieren reflow at 4 sampling stepsfew-step sampling · 4 sampling stepsDeSync: 0.917 -> 0.536Improved
SALSA-V instead of Frieren reflow at 4 sampling stepsfew-step sampling · 4 sampling stepsFAD_VGG: 2.17 -> 1.19Improved
SALSA-V instead of LoVA for 30 s generation (outpainting)long-form generation · 30-second long-form samplesDeSync: 1.225 -> 0.526Improved
SALSA-V instead of MMAudiohuman study · Human study MOSSynchronization: 2.92 -> 3.52Improved
SALSA-V instead of MMAudiohuman study · Human study MOSQuality: 3.16 -> 2.96Hurt
reduce sampling from 32 to 4 steps with shortcut modelSALSA-V · Generation time, 10 s cliptime (s): 4.84 -> 0.71Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy