AI avatar video platform with 240+ avatars and 140+ languages. January 2026 update added two major features: paragraph-level speech regeneration (fix one line without redoing the entire scene) and voice-speed controls (0.8x–1.2x globally or per scene). Speed controls work across all voice providers including ElevenLabs, Google, IBM, and Azure. PowerPoint-to-video conversion and interactive video features included. The quality ceiling just got high enough for customer-facing content.

AI voice generation platform with realistic text-to-speech, voice cloning from minutes of audio, 29+ language support, and HIPAA-compliant enterprise features. May 2026 SDK update (May 12–13) added voice metadata moderation, workspace API analytics for tracking voice production costs by client, and new LLM provider options inside ElevenAgents — including GPT 5.4 support.

AI-powered video and podcast editing now with an open beta API (April 13, 2026) that makes it chainable to Claude, GPTs, Zapier, and Make — enabling fully automated video workflows. The 'video waterfall': record → Descript auto-transcribes and edits → OpusClip extracts shorts → all without manual touches. Claude Opus 4.6 powers Underlord: B-roll accuracy 60%→92%, filler word removal +43%. Important: Descript is an editor, not a generator — you still need source footage. Teams chasing zero-footage promises will hit a wall. Entry plan $24/mo + Claude/GPT API costs typically $5-30/mo for SMB volume.