EdgeTX voice pack: make your radio speak in any voice
EdgeTX radios such as RadioMaster models announce switch positions, timers, telemetry and warnings with recorded voice files. The official packs are good, but many pilots want their radio to speak in an anime or game character’s voice, a deep announcer voice or their own.
This guide shows how the EdgeTX sound packs are organised, how to voice every phrase with the free SuperClonica engine and how to convert the files to the format EdgeTX plays.
How EdgeTX voice packs work
According to the EdgeTX sound pack repository, each language is a folder under SOUNDS on the SD card, for example SOUNDS/en, and you choose the voice language in the EdgeTX settings. The repository lists every phrase in a table per language, voices/en-US.csv, with columns for the text, the Path and the Filename, such as SYSTEM/0000.wav. The official packs themselves were generated with text to speech services.
Voice every phrase
Download the phrase table for your language, install the free engine with python -m pip install superclonica-cpu and, in Python, after import csv, os, from superclonica import Superclonica and tts = Superclonica(), then for row in csv.DictReader(open("phrases.csv", encoding="utf-8")): os.makedirs(os.path.join("raw", row["Path"]), exist_ok=True); audio, rate = tts.synthesize(row["Translation"], voice="voice.json", lang="en", seed=5); tts.save(os.path.join("raw", row["Path"], row["Filename"]), audio). Use a fixed seed so every phrase sounds like the same speaker. Read through phrases with numbers and units before you start, and write them the way they should be spoken.
Convert to the EdgeTX format
The repository describes the files as 16-bit PCM WAV, mono, up to 32 kHz. Convert each file with ffmpeg, for example ffmpeg -i raw/SYSTEM/0000.wav -ar 32000 -ac 1 -sample_fmt s16 SOUNDS/en/SYSTEM/0000.wav, keeping the same folder and file name. Copy the language folder to SOUNDS on the SD card, replacing the old one after a backup, then select the language in the radio settings.
Custom sounds for switches
For your own announcements, such as “Turtle mode on” or “Landing gear down”, voice the phrase, convert it the same way and copy it into the language folder. Then pick it with Play track in Special Functions or Global Functions on a switch. Keep phrases short so they finish before the next event.
Choose the voice
Try voices free on their pages with a phrase of 30 to 100 characters, no account needed: anime and game characters for a fun pack, announcers for clear callouts in the air. Make sure warnings stay easy to understand over motor noise. The engine speaks 31 languages, including Japanese, Korean and Russian; Chinese is not supported yet. Character voices in the catalog are fan-made, so share packs as fan creations.