Skip to content
Drix10 Blog

Riverside crosses $100M ARR milestone

, 4 items in AI Generated Music and Audio, 2 min read

In this digest (4 items)

Riverside announced it has crossed $100M ARR. The post lists 2.3 million creators, 16M+ guests, 40M+ conversations, and that 42% of Apple’s top 100 podcasts use Riverside.

Key points

  • ARR: $100M reached

  • 42% of Apple’s top 100 podcasts use Riverside

Sources

Neural Thickets – random logit bias matches weight perturbations gains

Neural Thickets perturbs every weight of a language model, keeps the best copies on a small labeled set, and lets them vote. A random bias added only to the output logits, leaving all other weights unchanged, matches or exceeds those gains across a diverse set of tasks.

Key points

  • Random bias on output logits, with all other weights unchanged, matches or exceeds gains from full weight perturbations.

  • Neural Thickets selects copies that score best on a small labeled set and uses voting.

Sources

Lumens LightPads launch with 16 pads, zero wiring

Lumens LightPads are now live, offering a Halloween-themed lighting experience. Tap a pad to trigger Hue lights, with 16 pads and no wiring required.

Key points

  • Product: Lumens LightPads are live.

  • Feature: 16 pads, zero wiring.

Sources

SingSync entry for open-source challenge released

SingSync is an entry for the open-source challenge that generates a potential full-length music video from one character image, one scene image and one sentence. It is built on ComfyUI using open-weight models and MIT-licensed code.

Key points

  • Built on ComfyUI, open-weight models, MIT-licensed code.

  • One character image, one scene image, one sentence can produce a full-length music video with an original song.

Sources