Synthesia voiceover: use your own AI voice
Synthesia makes training and explainer videos with AI avatars that read your script in one of its stock voices. Sometimes the stock voices are not what you want: you need your company’s narrator, your own voice or a character, or the same voice across tools.
This guide sums up Synthesia’s plans according to its pricing page and shows how its Voiceover Uploads feature lets you replace the text script with your own audio, made for free with SuperClonica.
Synthesia plans and minutes
On its pricing page in October 2026, the free Basic plan gives 10 minutes of video a month with a Synthesia watermark. Starter costs $19 a month, or $14 a month billed yearly, and adds about 12 minutes a month; Pro costs $89 a month, or $59 billed yearly, and adds about 60 minutes. Enterprise has custom pricing and unlimited minutes. The page lists more than 140 languages and voices.
Voiceover Uploads
Synthesia lists Voiceover Uploads, which lets you upload your own voiceover in place of a text script, so the avatar speaks your recording. Check which plan includes it on the pricing page. This is the way to give an avatar a voice Synthesia does not have, or to keep one narrator across all your videos and tools.
Make the voiceover offline
Write the script scene by scene, so each file matches one scene of the video. Install the free SuperClonica engine with python -m pip install superclonica-cpu and run superclonica --text-file scene-01.txt --voice my-voice.json --lang en --normalize --output scene-01.wav for each scene. It runs offline on an ordinary processor with no character limit, in 31 languages, and --normalize keeps every scene at the same loudness. Then upload the files in Synthesia.
Which voice to use
Hear catalog narrators and announcers free on their pages: paste 30 to 100 characters, no account needed. For a company narrator or your own voice, record a clean phrase of 5 to 10 seconds of a speaker who agreed and train it in the studio; training on your own NVIDIA graphics card is free. The trained voice speaks all 31 languages, which helps when one course goes out in several languages.
Commercial use
The ten stock voices of the engine, M1 to M5 and F1 to F5, can be used commercially within the OpenRAIL-M model license, and so can a voice trained from a speaker who agreed. Character voices in the catalog are fan-made and are not for company videos.