Operating Hours: Monday to Friday, 9am to 5pminfo@chakralabsai.com+94 72 063 5341fin
A white humanoid robot studying a large curved screen with a glowing neural network graph and a blue waveform in an AI research lab

What we are learning

Research

Papers and white papers from our team on Sinhala speech AI, phone voice pipelines and speech data.

  1. White paper Sept 2026 Seniru Epasinghe, Tharindu Sathsara, Dilanka Kasun, Sahan Dileepa, Bhawanthi Weerasinghe

    Recording Farmers' Voices: A Data Collection Standard for Sinhala Agricultural Speech

    The recording and transcription rules we set to collect clean, natural farmer speech, from silence padding to how numbers are written.
    • Speech Data
    • Dataset Design
    • Sinhala
    • Agriculture
    • Data Quality
    • Transcription
    First page of Recording Farmers' Voices: A Data Collection Standard for Sinhala Agricultural SpeechRead more
  2. Paper Sept 2026 Seniru Epasinghe, Tharindu Sathsara, Dilanka Kasun, Sahan Dileepa, Bhawanthi Weerasinghe

    Building a Sinhala Voice Pipeline for Phone Calls: From 3 to 8 Calls per GPU

    How we connected our own Sinhala speech-to-text, a language model and our own text-to-speech into a phone assistant that answers in about five seconds, and how many calls it can carry.
    • Voice Pipeline
    • Conversational AI
    • Sinhala
    • Telephony
    • Latency
    • Performance
    First page of Building a Sinhala Voice Pipeline for Phone Calls: From 3 to 8 Calls per GPURead more
  3. Paper Sept 2026 Seniru Epasinghe, Tharindu Sathsara, Dilanka Kasun, Sahan Dileepa, Bhawanthi Weerasinghe

    Teaching a Machine to Speak Sinhala: Training a Text-to-Speech Voice

    How we built a clear Sinhala voice for phone calls, and the speed-versus-clarity test that set it to 16 generation steps.
    • Speech AI
    • Sinhala
    • Text-to-Speech
    • Voice
    • Telephony
    • Low-latency
    First page of Teaching a Machine to Speak Sinhala: Training a Text-to-Speech VoiceRead more
  4. Paper Sept 2026 Seniru Epasinghe, Tharindu Sathsara, Dilanka Kasun, Sahan Dileepa, Bhawanthi Weerasinghe

    Training a Sinhala Speech-to-Text Model: Data, Training and Results

    How we trained Chakra's Sinhala speech-to-text on public recordings plus data we built and verified ourselves, cut the error rate, and tested it on real spontaneous speech.
    • Speech AI
    • Sinhala
    • Speech-to-Text
    • Model Training
    • Datasets
    • Low-resource Languages
    First page of Training a Sinhala Speech-to-Text Model: Data, Training and ResultsRead more

Interested in working together on research?

We are open to joint studies with universities, companies and public bodies. Tell us your idea.