Suno Speech beta generates narration and backing music in one track
First model to fuse narration with original scoring opens to everyone — plus Grok silently upgrades transcription and Tavus clears a video Turing test.
Copy markdown
Voice and music, generated as one track
Suno's Speech beta is the first model to generate narration and its own backing score together, not stitched from a separate TTS pass and a music track. It's open to everyone on Suno now — point it at a meditation, bedtime story, pep talk, or ASMR script and get a finished, scored read in one shot.
Grok retires its old transcription slug
xAI killed grok-voice-transcribe-1.0 today; every call to that slug now routes to grok-voice-transcribe-2.0 at the same price with higher accuracy. If you're still hardcoding the 1.0 slug, you're already on the better model — no code change, same bill.
Claude Code tightens auto-mode and Artifact permissions
Claude Code 2.1.287 narrows what auto-mode and Artifacts can do without a prompt, and fixes a Windows PATH bug that broke plugin-heavy installs. Worth the upgrade if you run Claude Code unattended, or on Windows with a stack of plugins.
Real-time video agents edge past human
Tavus reportedly ran its full-duplex Griffin-Lite through a video Turing test and 48% of participants thought it was a real person, up from 2.4% for the prior Phoenix model, at ~320ms latency. It's still gated to select testers in safety review — not an API you can call yet, but the clearest sign that real-time, human-grade video agents are close.