Big week. We rebuilt the podcast creation flow from the ground up, added an analytics dashboard to see how your shows are performing across platforms, and fixed a long list of issues in social clips, cover art, and podcast directory publishing. If something felt rough before, there's a good chance it got better.
New Features
Redesigned Podcast Creation Wizard
Creating a new podcast is now a guided, step-by-step experience instead of a single overwhelming form. The new wizard walks you through the essentials — basic info, review, configuration, host selection, and a celebratory completion screen — with your progress automatically saved along the way.
The biggest new trick: paste in your website URL and the wizard will analyze it, discover your key pages and any existing RSS feeds, and use that content to seed your podcast's knowledge base. From there, AI generates a podcast name, description, topics, and tags based on what it found — all reviewable and editable before you commit.
You can also pick your podcast hosts directly from the talent library during setup, with a searchable, filterable picker that lets you audition voices before choosing.
Analytics Dashboard for Admins
Admins now have a dedicated analytics page that pulls in distribution data from Mato. You can see downloads and unique listener counts across configurable time periods (7 days, 30 days, 90 days, or all time), with sparkline charts showing trends over the last two weeks. The dashboard also breaks down your top listening apps and countries, and shows a sortable table of all your shows with their episode counts and performance numbers.
Apple Podcasts Subcategory Support
You can now choose a subcategory when setting up podcast directory listings for Apple Podcasts. The subcategory picker automatically shows the relevant options based on your selected category, using Apple's official taxonomy. This gives your show a more precise placement in the Apple Podcasts directory, which can help listeners find it through browsing.
Host Avatars in AI Cover Art
When generating podcast cover art, the system now loads your hosts' profile photos and passes them to the AI as visual references. The result is cover art that actually reflects your hosts' appearance — facial features, skin tone, and hair — rather than generic illustrated figures. Host avatars take priority in the image budget, with brand reference images filling the remaining slots.
Retry Failed Podcast Creation
If the podcast creation wizard hits an error partway through, you no longer have to start over from scratch. You can now retry from wherever it failed — whether that's generating hosts, filling in podcast fields, or the final creation step — without losing any of your previous inputs.
Improvements
Platform Icons Link to the Right Places
Platform icons on public podcast pages (Spotify, Apple Podcasts, YouTube Music, and others) now correctly link to each platform's listing page for that show. Previously they were all falling back to the RSS feed URL. Icons for platforms where no listing URL is available now appear dimmed rather than linking somewhere unhelpful.
Directory Status Updates Now Work for All Platforms
Updating directory listing status in your podcast's publishing settings was silently failing for YouTube Music, Apple Podcasts, Amazon Music, and iHeart Radio. These platforms use slightly different names in the API versus the database, and the mapping was only working in one direction. That's fixed — status changes now save correctly for all platforms.
Talent Personality Card Is Much More Readable
The personality data shown on talent profiles was displaying raw JSON for nested objects and showing "[object Object]" for anecdotes. These now render as proper formatted key-value pairs and styled quote cards. Short lists of two items or fewer are shown inline; longer lists become clean bulleted lists.
AI Personality Syncs No Longer Wipe Episode History
When you used AI Auto-Generate to refresh a host's personality, it was silently clearing out accumulated episode and story data from the host's memory. The sync now reads existing memory blocks first and preserves any non-empty content history and anecdotes before writing new data.
Social Clips Colors Are Smarter and More Reliable
The AI that picks colors for social video clips now uses a vision-capable model and receives your podcast's cover art as input, so the generated color palette matches your show's visual identity. Several bugs were fixed along the way: highlighted caption words were sometimes invisible because the highlight color was identical to the active text color; user color edits were getting lost on save and reload; and some templates were using the wrong fallback colors. All of these are resolved.
Caption Display in Social Clips
Words in social clip captions now display in natural groups rather than one at a time or all at once. This matches the TikTok-style caption behavior most viewers expect — a phrase appears on screen, the active word is highlighted, then it advances to the next group. The density of word groupings is also now correctly wired to the editor's caption density control.
Talent Audio Samples Generate Reliably in Production
Generating voice preview samples for hosts was failing in production with a credentials error. The fix consolidates credential handling into a shared factory, and restores the correct single-speaker synthesis approach for samples (the previous version was using the multi-speaker dialogue pipeline by mistake, which produced corrupted audio files).
More Reliable Background Article Filtering
The article filtering job was occasionally failing with a size limit error when processing large sets of articles. The fix reduces what gets passed between job steps to lightweight references only, then re-fetches full content as needed. This prevents the job from hitting processing limits and keeps article scoring working correctly.
Bug Fixes
Fixed an error in episode generation where the segment schema was causing structured output failures with OpenAI. The schema was using array defaults in a way that OpenAI rejected, and affected fields were being omitted from generation outputs.
Fixed database query errors that were causing intermittent failures across content sources, talent assignments, and RSS feed fetching. These were caused by ambiguous foreign key references after a new relationship was added to the podcast feeds table — all affected queries now specify the correct relationship explicitly.
Fixed display of talent personality traits that included nested data structures — patterns, voice characteristics, and similar fields now render as readable formatted content instead of raw code.