BATS - Speech2Speech

Also available in:

Deutsch

Speech in any language? No – in English: transcribe, translate, re-voice

Speech2Speech turns spoken audio in any Whisper-supported language into an English voice-over: the assistant-created entry transcribes the audio with Whisper (model large-v2) and translates the text to English on the way – then the AI voice engine Kokoro (Nvidia GPU) speaks the text in a choice of US or UK English voices. A foreign-language interview or voice-over becomes an English-language audio clip with the same timing.

Adding the menu entry

“Speech2Speech” is added through the “Add function...” assistant (right-click in the BIN → Add function... → pick Speech2Speech - Create English Voice). The assistant’s questions:

  1. Voice (“Please select the voice. English only.”): Heart (American, female), Michael (American, male), Lily (British, female) or George (British, male).
  2. Max. stretch (0–50 %, default 10 %): speech-rate adjustment to fit the sentences into the original time slots; recommended below 15 %, 0 % = off.
  3. Add result to the BIN? (“this preset will use the trimmed version of the selected file”) – Yes → additionally registered under __BAT Server\Speech.
  4. Placement (BIN or timeline) and name (suggestion: “Translate by <voice>”), then [Preview '.bat0' file] / [Add menu entry].

How to use it

  1. Right-click an audio clip (or a clip with audio) → Generate Speech submenu → your Speech2Speech entry (e.g. “Translate by Heart”).
  2. BAT Server extracts the audio of the IN/OUT range (16 kHz mono), transcribes + translates it with Whisper-Faster (large-v2) into an English SRT …
  3. … and lets Kokoro speak the SRT in the chosen voice. The WAV file is registered as an audio clip in the BIN (additionally under __BAT Server\Speech if you chose so in the assistant) – time stamps included, matching the clip’s frame rate.

Good to know

  • The result is always English: the source may be in any Whisper-supported language (the language is auto-detected) – the output voice is US or UK English.
  • Timing as in the original: the transcribed sentences inherit the time slots of the original speech; the stretch limit controls how far the speech rate is adapted.
  • Preconfigured on board: BAT Server additionally ships the system preset “Translate to English” (Whisper → Kokoro, voice Heart) for audio clips in the BIN – the assistant is for your own voices, names and placements.
  • Nvidia GPU: Kokoro runs on the Nvidia GPU; without a suitable GPU the entry is not really usable.

Managing the preset

Entries created via the assistant are user presets: the .bat0 file lives under C:\ProgramData\EdiusPowerTools\BAT-server\bat_files\BIN\ or …\bat_files\timeline\ depending on your choice, and can be freely edited, renamed or deleted (deleting removes the context-menu entry). Voice and stretch value are baked into the .bat0 – changes take effect from the next run.