A lot landed this week. We shipped a visual calendar for scheduling episodes, video recording inside live interviews, automatic highlight clip generation, and a brand-new Help Center with step-by-step guides for every major workflow. There's also a critical fix that unblocked podcast generation for some accounts, plus a long list of reliability and consistency improvements across billing, AI generation, and the audio pipeline.
New Features
Episode Scheduling with Drag-and-Drop Calendar
You can now plan your episode releases on a visual calendar instead of managing a flat queue. The schedule page offers four different view combinations — week or month, list or calendar grid — so you can see your content pipeline in whatever format makes sense for you.
Drag suggestions from the queue onto any date to schedule them, swap two scheduled episodes with a single drag, or unslot them back to the queue. There's also an inline settings panel for configuring your publishing frequency, time, timezone, and active days. Scheduled episodes take priority in the generation queue, and overdue entries get automatically caught up, so nothing falls through the cracks.
Video Recording During Interviews
Interviews can now capture video alongside audio. When video recording is enabled, a recording indicator shows the current duration so you always know you're still capturing. A series of reliability fixes landed at the same time — including a fix for audio that was getting cut off before the recording fully saved, more consistent session persistence, and better state management throughout.
Automatic Interview Clip Generation
After an interview wraps, Mato can now analyze the session and automatically generate highlight clips from the best moments. The clip selector scores each segment of the interview and extracts the top moments into short-form video clips, complete with captions. This is Phase 1 of the clips pipeline — guest-facing clips with the AI host avatar integration coming separately.
Help Center is Live
The Mato Help Center launched this week with 15+ articles covering every major part of the platform. New guides walk through the full onboarding flow, creating your first podcast, setting up hosts and voices, adding content sources, generating and editing episodes, the script editor, metadata management, publishing, settings configuration, branding, Mato Distribution setup, player embedding, creating custom hosts, and voice configuration.
Articles are synced automatically to Intercom so they're available from anywhere inside the app.
Improvements
Podcast Generation Now Defaults to Anthropic
All new podcasts now use Anthropic as the default AI provider for script generation. This also resolves a production issue where some accounts were hitting generation failures because OpenAI's structured output format requires a stricter schema shape that the previous default configuration didn't satisfy. Existing podcasts that explicitly chose OpenAI will continue to use it, and they'll also benefit from the schema fix.
Billing Plan Cards Are Now Consistent
The plan cards shown during onboarding and on the team billing page are now powered by the same shared configuration. Plan names, prices, and layout are identical across both surfaces. We also added upfront validation to catch misconfigured pricing setups before sending users to Stripe, so checkout failures surface earlier with a clear message rather than deep inside the payment flow.
Intercom Deep-Links Into Your Active Podcast
Intercom messages, checklists, and tours can now link directly into the right place in your workspace using short static URLs like /go/my-podcast or /go/create-episode. The system tracks which podcast you were most recently working on and routes you there automatically — so a "Set up distribution" checklist step lands on the podcast you actually care about, not a random one.
Audio Processor Secured and Streamlined
The Modal audio processing service was consolidated from ten separate endpoints into a single unified endpoint. This freed up capacity to deploy the new video extraction service. The consolidated endpoint now requires authenticated requests, so the processing pipeline is more secure from arbitrary external calls.
Pipeline Failures Are Now Caught Automatically
Episodes and background jobs that get stuck — because of memory pressure, unexpected process termination, or other failures — are now automatically detected and cleaned up. A background monitor runs every 15 minutes and flips any jobs or episodes that have been in an in-progress state too long to a failed state, so they stop blocking the queue. All pipeline failures are also now captured in a structured log, making them much easier to investigate.
Bug Fixes
Podcast Generation Was Failing for Some Accounts
A schema incompatibility between the claim provenance format introduced in recent weeks and OpenAI's strict output mode was causing generation to fail at the first script segment for any account using OpenAI as the AI provider. The fix introduces a compatible schema format for generation output while leaving the existing parsing logic untouched.
Help Center Articles Were Not Appearing on the Website
Several batches of published Help Center articles were marked as drafts in their content files, which caused them to be silently excluded from the documentation listing on the website even though they had been synced to Intercom. All affected articles — covering episode workflows, podcast settings, distribution setup, and talent management — are now correctly published and visible.
Engagement Email Job Was Crashing
The AI model identifier used in the engagement topic approval email job contained a formatting error that caused runtime failures every time the job ran. The correct identifier is now in place and the job processes normally.
Help Center Images Were Failing to Sync to Intercom
The Intercom sync pipeline was failing when articles used absolute image paths, sending Intercom a URL it couldn't resolve. The sync now handles all three image path formats correctly and waits for screenshot assets to be publicly available before pushing an article live, preventing intermittent failures on fresh deploys.