Speech Recognition Jobs

I’m gathering a small set of natural Amharic conversations to improve an AI speech-recognition model. I only need recordings from a handful of native speakers—no more than ten in total—so the project stays personal and manageable. Each conversation should revolve around ordinary daily activities: cooking, shopping, commuting, family routines, leisure time, or any other day-to-day topic that feels authentic to you. Think of it as a relaxed chat you might have with a friend or relative; the more genuine the flow, the better the data will train the model. Audio quality matters. Please record in a quiet space, use a decent microphone (a modern phone headset is usually fine), and save the files in WAV or high-quality MP3. Minimal background noise and clear diction are es...

This project focuses on quality assurance (QA), transcription verification, and audio segmentation for Czech-language recordings. Contributors will review audio recordings, verify transcription accuracy, identify annotation errors, and ensure that audio segments are correctly divided and aligned with the spoken content. Key Responsibilities Audio Quality Assurance: Review Czech audio recordings and validate existing transcriptions and annotations. Transcription Verification: Identify and correct errors in spelling, punctuation, word usage, and transcription accuracy according to the provided guidelines. Audio Segmentation: Divide recordings into appropriate segments based on speech boundaries, ensuring accurate start and end timestamps and proper alignment between the audio and its cor...

I’m preparing a long-term AI speech data project and need native Czech linguists who can handle full-cycle audio annotation. The tasks on each recording include: transcribing the Czech speech verbatim, segmenting sentences, inserting precise timestamps, tagging speakers, and labeling background noise as well as any overlaps. Once your pass is complete, you’ll also run a quick self-quality check so every delivered hour is clean and consistent. We will begin with a short pilot batch. It lets me gauge accuracy and lets you get comfortable with the guidelines before we scale to larger volumes. I pay USD 75 for every audio hour that passes our internal acceptance review. To move forward, please reply with: • a brief outline of your previous Czech transcription or annotation e...

I’m kicking off a pilot to build a high-quality Czech speech corpus and need a meticulous native speaker to help me. Your first task will focus on Czech audio transcription of speeches or presentations, followed by sentence segmentation, speaker identification, precise timestamps, and detailed noise/overlap labelling. Every hour that meets our quality bar is paid at USD 75; once the pilot proves accurate, we’ll expand to a longer-term collaboration. Deliverables for the pilot • Fully transcribed Czech audio (speeches/presentations) in UTF-8 text • Sentence-level time-aligned segments with speaker tags • Noise and overlap markers applied to guidelines I’ll supply • Quick self-check report confirming the annotation passes our QC checklist To mo...

I need open-source speech recognition stacks fine-tuning for high accuracy, reliable, fully on-prem deployment across seven Indian languages—Hindi, Tamil, Bengali, Marathi, Telugu, Kannada and Gujarati. My current shortlist of engines is Whisper, AI4Bharat’s Indic models and Vanni from IISc; you are free to combine or layer these as long as the final pipeline remains 100 % open source and runs locally on our GPU servers (Windows CUDA, PyTorch). What I want from you is a measurable lift in word-error-rate, inference speed and resource efficiency. Techniques such as transfer-learning, domain-specific vocab augmentation, quantisation, batching or ONNX/TensorRT optimisation are all on the table—choose what works best for each language pair. Deliverables • A repro...