Fluid Intelligence
An optional model that cleans each line of the transcript. Altic licenses it for personal, non-commercial use only.
Before you start
Caution: Do not use Fluid Intelligence in a church or other organization without written permission from Altic. Altic licenses the model for personal, non-commercial use only.
What Fluid Intelligence does
FlowCast can send each line of the transcript through Fluid Intelligence, the model that FluidVoice uses to clean dictation. The model adds punctuation and capitals, and it corrects some words that FlowCast heard incorrectly. It can also change a correct word, such as “spake” to “spoke”. The model operates on the computer.
Fluid Intelligence is off until you turn it on.
The model
On a Mac, if FluidVoice downloaded the model, FlowCast uses that copy. If not, FlowCast can download the model (3.6 GB).
On Windows, FlowCast downloads the model as a GGUF file (3.4 GB) and operates it on the CPU with llama.cpp.
The model uses approximately 3 GB of memory while it is on.
Turn on Fluid Intelligence
- Click the Settings button in the top bar.
- Click Transcript.
- If the model is not on the computer, click Download.
- Turn on Fluid Intelligence.
How the corrected lines operate
At the end of each line, the corrected line replaces the line that FlowCast heard. A reference in the heard line goes live immediately, and the corrected line does not change it. A reference that only the corrected line gives goes live if the heard line gave no reference. The Detected panel compares the corrected words with the verses.