Scratch voices for animatics and storyboards
An animatic without sound is hard to judge: you cannot feel the timing of a joke or the length of a pause until someone says the line. Scratch tracks are usually recorded by whoever is around, and every script change means recording again.
SuperClonica turns script lines into scratch dialogue in seconds, with a different voice for every character, free to test online and offline on your own computer. This guide shows how to set it up for an animation pipeline and where it should stop.
Why AI scratch voices help
Scratch dialogue is a timing tool; the performance comes later, from actors.
- Timing: hear how long each line really takes before you draw a single in-between.
- Rewrites: change a word in the script and the line is ready again a second later.
- Consistency: the same voice and the same --seed give the same take every time.
- Variety: a different voice for every character, from narrators to cartoon voices.
Try a character line free online
On the page of any voice, type a line of 30 to 100 characters, choose the language and press Voice it. In a few seconds you can hear whether the voice fits the character and download the take, with no account.
Set up the lines
The ten stock voices of the model, M1 to M5 and F1 to F5, are free and cover a basic cast; the catalog adds narrators and character voices.
- One text file per line, named by scene, line and character, such as sc04_line07_mira.txt.
- Punctuation for timing: an ellipsis for a hesitation, a full stop for a beat, a question mark for a rising line.
- One voice per character for the whole project, written down in a cast list.
- A fixed --seed per character, so a re-voiced line matches the earlier takes.
Voice the whole scene offline
Install the free SuperClonica engine: python -m pip install superclonica-cpu. Voice a line with one command: superclonica --text-file sc04_line07.txt --voice F4 --lang en --seed 3 --output sc04_line07.wav. The 16-bit WAV goes straight into Blender, Toon Boom, Premiere, DaVinci Resolve or any editor. For a whole scene, the Python API voices a list of lines with a voice for each, in order, with tts.synthesize_batch(), and --speed tightens or slows a line without rewriting it. It runs offline on an ordinary processor, so unreleased scripts never leave the studio.
Scratch is not the final cast
Use scratch voices inside the team, for timing, pitches and test screenings. Character and parody voices from the catalog are fan-made, so keep them out of anything you publish. For the final film, cast actors; if you ever want a synthetic version of an actor’s voice for pickups, get their written consent and agree on payment first. For your own original characters, a voice trained from a performer who agreed can carry from animatic to release.