Building MemeVibe: Putting Someone Inside a Trending Video From a Single PhotoEvery week a new short video format takes over the feed, and everyone rushes to remake it. Recreating one yourself usually means filming, editing and nailing the timing, and by the time you're done th7 min read·peters-tech-lab.hashnode.dev5d ago
Two People Talking Is a Different Problem: Adding Speaker Diarization to an AI Dubbing PipelineSingle-speaker dubbing is a solved pipeline. The moment a second voice shows up, four of the five stages break in different ways. For the first several months, AI Video Translator assumed one speaker7 min read·peters-tech-lab.hashnode.devNov 6
CBZ, CBR, EPUB, PDF: What I Learned Getting Manga Page Order Right Across Six File FormatsWhen I started building AI Manga Translator, I assumed the hard part would be OCR and translation. Then users started uploading whole volumes, and I discovered a much more boring problem that broke th6 min read·peters-tech-lab.hashnode.devNov 1
Soft Subtitles vs Burned-In: Why Our AI Dubs Now Ship a Separate SRT FileGenerating captions and dubbed audio from one reviewed script, and why that matters more than the export format. This week I shipped three updates to AI Video Translator: line-by-line review of transl6 min read·peters-tech-lab.hashnode.devOct 21
Slicing a 30,000px Webtoon Without Cutting Speech Bubbles in HalfA practical approach to splitting very tall images for OCR and translation, with code.6 min read·peters-tech-lab.hashnode.devOct 7