← NeuralDrift Weekly

ISSUE #1 · 2026-09-15

ComfyUI 0.35 Lands — Plus New Local Video & Audio Tools

ComfyUI 0.35 shipped upstream, NeuralDrift has current RTX 5080 execution records for AnimateDiff and ACE-Step audio workflows, and three local video/audio projects are worth watching.

ComfyUI 0.35 landed upstream

Upstream release — not independently execution-tested

ComfyUI 0.35 shipped this week. As with any point release, node behavior and default settings can shift under existing graphs, so treat this as upstream release news rather than a NeuralDrift-verified compatibility statement.

We haven't re-run every catalog workflow against 0.35 yet. The execution records linked below reflect the ComfyUI and PyTorch versions actually used at test time — check the "Testing & source information" section on any workflow page before assuming a new release changes nothing.

Local video: AnimateDiff stays current on RTX 5080

NeuralDrift-tested — RTX 5080, current evidence

Five AnimateDiff graphs in the catalog — ambient motion, loop-oriented, landscape drift, product rotation, and slow zoom — have current, successful execution records on an RTX 5080. These are the workflows behind this issue's "local video tools" claim; nothing here is a catalog estimate.

Each record lists the settings and environment used for its own run. Use those records to assess fit for your hardware rather than treating the five graphs as a VRAM ranking.

Local audio: ACE-Step 1.5 turbo

NeuralDrift-tested — RTX 5080, current evidence

The ACE-Step 1.5 turbo checkpoint graph — generating audio from musical tags and lyrics — also has a current, successful RTX 5080 execution record in the catalog. Pair it with the ACE-Step 1.5 guide in Guides if you're setting it up for the first time.

Being watched: SolarWM, AuK, FireRedAudio

Community reports — not run or verified by NeuralDrift

Three more local video/audio projects came up this week in community discussion: SolarWM (local video/world-model tooling), AuK, and FireRedAudio (an open local text-to-speech/audio project). None of these have been installed, run, or execution-tested by NeuralDrift yet.

We're listing them because they're relevant to this issue's theme, not because we're vouching for their claims. If we run them locally and get a real result — success or failure — it'll show up in the Lab with an execution record, the same as everything else on this site.

GO DEEPER

// NeuralDrift Weekly

Get NeuralDrift Weekly

Execution-tested workflows, compatibility findings, ComfyUI changes, and practical local-AI guidance — delivered weekly.

No spam, unsubscribe anytime. See our privacy policy.