Customize speech to text for the best results
For customers and partners willing to invest in their own speech recognition models we offer customization. At Scriptix we build all our speech to text models in house, meaning together with our partners and customers we can train models on their data. Generic language models can work in many cases, but for more specific use cases customization is the way to go. Think about specific verticals like politics, health, sports, and law.
Key features
a customized model can reach as much as 15% higher accuracy than generic models.
we build the models for a specific customer or partner and host it privately.
your model will be made available to you on our API-platform to be integrated in your workflow.
we can deliver a custom model within 2 months and run continuous updates.
Customization: how does it work?
Get in touch
For customization, every use case is different. Depending on the training data available necessary to train a custom model we determine cost.
Share data
To develop a custom model Scriptix requires data from its customers and partners. Depending on how much customization is required, on average 1.000 hours of audio and 100 hours of perfect transcripts or subtitles are needed to train a custom model.
Process content
Once done, your custom model will be hosted privately to your account and as such become accessible only to you. Use the model either in batch or real-time on our cloud platform or hosted to enrich your products and add more value for your customers.
Requested by large broadcaters, streaming providers, call centers and system integrators
- Hosted privately
- Cloud or on prem
- Cost depending on data available
- Exclusivity: only your organization can use the model
