PodcastorAI
Your AI twin hosts your video podcast
720 followers
Your AI twin hosts your video podcast
720 followers
Create studio-quality video podcasts from scripts, links, PDFs, or ideas — with AI hosts, natural voices, and your digital twin. Podcastor handles production and puts you on screen, turning hours of recording and editing into a finished video podcast in around 15 minutes. Create solo or two-host episodes and get a publish-ready video without cameras or a production team. Already have an audio episode? Upload it and transform it into a video podcast with your voice and AI avatar.
the digital twin part is the interesting bet - if you re-record one line weeks later does the AI voice/face stay consistent with the original episode, or is there drift between sessions?
PodcastorAI
@sylvialane makes sense that identity persistence was a deliberate design target rather than a side effect - that's the harder engineering problem than the initial generation.
That is exactly the balance I was getting at. Turning one full episode into platform-ready clips without adding a second review queue sounds like the right direction, especially if the original context stays attached to the cuts.
PodcastorAI
@wesc Exactly. A clip isn’t useful if it removes the nuance that made the original point valuable. Our goal is to make repurposing genuinely additive: preserve the speaker’s intent and context while adapting each cut for the platform, without creating another editing workflow for the creator. That principle is guiding how we’re building the AI clipping feature.
PodcastorAI
@wesc Honestly Wes, this is the stuff that keeps me up at night. The second review queue is exactly the trap — you save time on the cut and lose it all again checking the clips. Keeping the original context attached is the part we're still wrestling with, but you just described the version we're trying to get to.
AskCodi
My favorite part isn't even the AI avatar, it's compressing the entire production workflow into minutes. That changes who can consistently create content. Curious, how much editing do users typically do after the first generated version?
PodcastorAI
@shreyans_assistiv Thanks Shreyans! 🙏 That's exactly the shift we were chasing — turning production from a multi-day process into something you can do between meetings.
On editing: most users keep the first generated version with only light tweaks — usually swapping a line or two of script and adjusting pacing. The people who edit more are the ones dialing in brand tone, and that's the workflow we're focused on making faster next.
PicWish
are you running custom models for avatar or external APIs? generating 30 min videos in 15 mins is good latency.
PodcastorAI
@mohsinproduct Thanks, Mohsin! We use a combination of in-house models and selected external video-generation models.
Our proprietary models are mainly focused on lip-sync and mouth-movement generation, while we integrate external video models for other parts of the production process.
The audio-upload path is the part that catches my eye. Most podcasters I talk to have a big back catalog of audio-only episodes and no realistic way to get them onto YouTube as video. Have you seen how the AI-avatar episodes hold up on retention versus real-camera footage? That's the number that would decide it for the audio-first crowd.
PodcastorAI
@podcast_ai That’s the right metric, and we don’t have enough controlled retention data yet to claim that AI-avatar episodes match real-camera footage. Before building Podcastor, we studied a number of AI podcast channels on YouTube with tens or even hundreds of thousands of subscribers and strong view counts, especially in storytelling and education. There’s still a gap versus great on-camera content, but those categories show that AI-hosted formats can already work well when the content fits.
We’re now running our own YouTube experiments across several topics and collecting more data. We’d love to have more audio-first creators join us so we can build a clearer picture of where this format creates the most value.
The line between useful and creepy here probably isn’t the voice clone. It’s whether the human still owns the point of view. If AI handles the production layer and the hosts keep the banter, I can see the appeal. What part of the workflow are you trying to remove first?
PodcastorAI
@ra5tadark Really well said, Rasul. We don’t want to remove the human point of view — that’s actually the part we think should stay.
The first workflow we’re trying to remove is the production burden: turning raw ideas or documents into a structured conversation, creating the host flow, and producing the audio/video output. The human should still own the perspective, taste, and final direction.
YourSitee
the dedicated study podcast flow makes the pdf input more than a creator feature: students can turn lecture notes, textbook chapters, slides, or study guides into audio or video for review. how much control do they get over depth and which sections the script emphasizes?
PodcastorAI
@szalovszky Great point, Dávid. That study use case is exactly one of the directions we’re excited about.
Right now, users can guide the output by choosing the source material and shaping the prompt/episode direction, but we want to give much more control over depth, tone, and which sections get emphasized. Section-level emphasis and “short overview vs deep dive” controls are high on our list.
Would love to hear what level of control would feel ideal for students.