v1.2.0

Added

• Podcast generation. Gotcha can now turn text into listenable audio. Ask
  for a topic, an article or your notes as a podcast and the assistant writes
  the script and speaks it through your configured text-to-speech API — as a
  single narrator (synthesize_podcast) or as two hosts in conversation with
  distinct voices (synthesize_podcast_dialogue; pick the host voices in
  Settings → Speech). The assistant paces the conversation as it writes it,
  choosing the silence after each turn — a quick interjection runs straight
  into the reply, a revelation gets a beat first — so an episode breathes
  the way the script intended instead of ticking along on a fixed gap.
  Long scripts are synthesized in segments and joined on the device, episodes
  land under Gotcha/Podcasts as .m4a (or .mp3 when Termux's ffmpeg is
  available), and share_podcast opens the system share sheet to send one
  anywhere. A voice memo can become an episode too: transcribe_file turns a
  saved recording into text — reading the file without ever modifying or
  deleting it — for the assistant to script and re-voice. Speech synthesis and
  transcription use your configured TTS/STT endpoints; the audio assembly never
  leaves the phone. Android's built-in TTS cannot write files, so these tools
  need Samosa AI or an External API selected under Settings → Speech.
• Audio format conversion. A new media_convert tool converts audio between
  MP3, M4A, AAC, OGG, Opus, WAV and FLAC using Termux's ffmpeg — still entirely
  on the device, with nothing uploaded anywhere. This is the only way to get an
  MP3 out of Gotcha: Android ships no MP3 encoder at all, so media_edit cannot
  write one however it is asked. The tool appears only when Termux is installed,
  and if its ffmpeg package is missing the error says exactly what to run.
• Audio & video editing. A new media_edit tool edits media entirely
  on-device: trim a window out of a clip, pull a video's audio into its own
  file, mute a video, shrink one to a smaller resolution, speed it up or slow it
  down, and join several files end to end, plus an info read for duration,
  resolution and tracks. Video and audio files work the same way, so trimming a
  voice memo and trimming a 4K video are the same request. Trimming, extracting
  and muting copy the streams rather than re-encoding, so they finish in seconds
  and lose no quality at all; compressing, re-timing and joining do re-encode,
  and Gotcha now says which of the two happened so you know whether a file was
  degraded. Every operation writes a new file and leaves the original alone.
  Adding music, overlaying text and editing what is inside a frame are not
  possible, and DRM-protected media cannot be opened at all — a bundled
  media_editing skill teaches the assistant to say so rather than attempt a
  workaround.
• PDF editing. A new pdf_edit tool reshapes PDFs entirely on-device: merge
  several files into one, split into single pages, extract or delete a page
  range, and rotate pages, plus an info read for the page count. Encrypted
  files open with a password you supply — but because the edited copy cannot keep
  that protection, Gotcha now refuses the edit until it has told you the copy
  will open without a password and you have agreed. Every operation writes a new
  file and leaves the original alone unless you ask for it to be replaced. Editing the
  text or images printed on a page is not possible — PDF stores glyphs at fixed
  coordinates, so there is nothing to reflow — and a bundled pdf_editing skill
  teaches the assistant to say so rather than fake it.