Low latency
Live captions
Accessibility
Voice command services
Chatbots
Virtual assistants
On-premise
Speech-To-Text models
State of the art WER.
Available languages
English
Dutch
French
Portuguese
Spanish
German
Italian
Ready soon:
HebrewBasic
Essential features for a solid start. Perfect for individuals and small projects.
12 500
hours included
250 $ / month
Contact us0.04$ / hour
Standard
Balanced features for a growing need. Ideal for businesses seeking a comprehensive solution.
50 000
hours included
750 $ / month
Contact us0.02$ / hour
Model Features
Ultimate Privacy. Your hardware. Data never leaves your servers.No GPU needed50x cheaper than Google ASR (starting at 0.01 $/hour)State-of-the-art WER (word error rate)x10 real-time on a single CPU core
Post-processed Models
Great for
Meeting transcripts - online and offline
Visual voicemail
Voice memo transcripts
Content creation
Productivity and analytics
Movie subtitles
x20 real-time on a single CPU core
Optimized for calls (8kHz or 16kHz)
Models optimized for Call Centers with highest accuracy, for processing recorded speech.
Websockets APIPunctuationWord timestampsCapitalizationNo hallucinationsFrequently Asked Questions
What are the server requirements for the Self-Hosted Speech-To-Text solution?
CPU-Optimized, No Hardware Requirements. Banafo’s ASR models can run on modest CPU. Banafo can reside on your laptop, no GPU needed!
Get speech-to-text into your apps and services hosted on
your servers
on your laptop
on your mobile phone!
Which languages does the ASR system support?
Bulgarian, English, Dutch, French, German, Hebrew, Hungarian, Italian, Japanese, Portuguese, Spanish
How is security ensured in the Self-Hosted solution?
These are on premise models and nothing passes through our servers. Data Privacy at its core. Your recordings are not shared with us or third parties and stay on your systems at all times. Your data is your most valuable asset, and we believe it should stay in your hands. With our self-hosted ASR models, you maintain complete control over your sensitive information, ensuring it resides in your infrastructure.
What is the expected accuracy of the ASR models?
These are the most accurate ASR models on the market* Unmatched Accuracy, Low WER. Even in noisy conditions or with accented speakers, Banafo performs flawlessly. Our ASR models are trained on vast datasets, ensuring industry-leading accuracy. Say goodbye to misinterpretations and hello to precise transcriptions and voice commands, enhancing user experience and overall efficiency.
How does the system handle speech hallucinations?
No hallucinations. Banafo’s ASR models do not make up words that were never said, and do not change the meaning. Banafo is meticulously crafted to understand and transcribe spoken language without the pitfalls of hallucinations, ensuring that your transcripts reflect the true essence of the spoken word.
Is the solution scalable, and how?
Scalability and Cost-Efficiency: With a Websockets based server license, scale your ASR capabilities based on your demands. Our self-hosted models are designed to be highly scalable, enabling you to handle an increasing volume of requests without compromising performance. You pay for the capacity you want to use!
How easily can the solution be integrated into existing systems?
Seamless Integration: Whether you're developing a speech-enabled application, virtual assistant, or visual voicemail, our self-hosted ASR models seamlessly integrate into your existing setup. Experience effortless integration with our developer-friendly models.
What is the processing speed of the Speech-To-Text models?
Speed Speech-To-Text Streaming 20 times real-time on a single CPU
Speech Speech-To-Text Post-Processed 10 times real-time on a single CPU
What is the pricing model for the Self-Hosted solution?
These are the lowest cost models on the market* Check the price here.