
Podcast & Live to Shorts: Two-Speaker Clips, PT/ES
Finished clips from your podcast, live or talk show, built for Portuguese and Spanish first. Two people side by side become one above the other, with the captions in the band between them. Word-by-word captions keep accents (ã, ç, ñ, ¿) and strong words can be masked. Every clip comes with a title, description, hashtags and a "Full episode" source line. 5 to 8 vertical MP4s per request.
#Video#Sociale media
Beoordeling
Meer beoordelingen nodig
Verkocht
0
Gebruikswijze
Uitvoeren op Capafy
Ook in externe apps
Publisher levert
Claude Sonnet 5
Cortes from your talk show, ready to post
Built for Portuguese and Spanish first. Three things it does for you:
- Two people side by side become one above the other. In a podcast filmed with one wide camera, each person gets their half of the vertical screen and the captions sit in the band between them, covering no one's face.
- Every clip comes with a post kit in the language of the video: a title that makes people want the rest without inventing anything, a description, hashtags and a "Full episode: " source line.
- Captions with the right accents (ã, õ, ç, é, ñ, ¿, ¡) and, if you ask, strong words masked (f***).
What you get for each clip
- A finished 9:16 MP4 at 1080x1920, for TikTok, Reels and YouTube Shorts
- Automatic framing: two people one above the other, or following whoever is on screen, or the whole picture over a blurred background. You can also choose
- Word-by-word captions burned in, proofread in context, plus the .srt file
- Level sound at the loudness the apps use
- The post kit in a text file per clip, plus a clips.csv with everything and a map of where each clip sits in the episode
What to send
- The episode: upload it under about 50 MB, or paste a link to the file in Google Drive ("anyone with the link") or Dropbox, up to 4 GB.
- A long episode: each request reads up to 60 minutes. Say which part (for example 1:10:00-2:10:00); another part is another request.
- The link to the full episode, for the source line of every clip.
- Only a public YouTube link: you get the clip plan with the post kits first (nothing is downloaded from YouTube). Send the file afterwards and the same clips come back as MP4s.
- Optional: how many clips (default 5 to 8, up to 10), the length, names to spell right, "no strong words".
What it will not do
- It is for your own show, or videos you have permission to edit. Nothing is downloaded from YouTube, TikTok or similar sites.
- It finds faces, not voices. It does not know who is talking. If the camera cuts between shots inside a clip, it uses the usual framing instead of splitting the screen.
- Captions come from speech recognition. Read each clip before posting. Masking changes the text only, not the sound.
- No promises about views.
What goes online
The first run downloads software from PyPI and the open speech model from Hugging Face. Your video and audio are not sent to either. The transcript text and check frames are read by the AI model running this agent. When you paste a YouTube link, only the link is sent to Google's Gemini API, which listens to that public video and returns the words with times.
Externe API’s
PyPI · Hugging Face · GitHub · User-provided file link · Google Gemini API

