Speech collection project for AI Development (DE, NOT NEEDED TO BE A LINGUIST)

• Task: Speech collection project within the context of Artificial Intelligence Development. This task involves collecting speech samples: reading the source data while adopting various emotional styles (about two hours, ten minutes per emotion). You will need to read the text with different tones to convey varying emotions such as sadness, anger and happiness (requested emotion will be always highlighted in the tool for recording). PLEASE NOTE THAT THIS IS NOT A TRANSLATION TASK.
• Language: German (all accents are acceptable)
• Requirements:
• Native speaker fluency in the target language. No linguistic background is required.
• Full name and email address of individual speakers as we need to assign the tasks individually to the resources. Rate related communication would be directed to the provided PM alias.
• Understanding training instructions written in English.
• A stable high-speed internet connection.
• A working microphone (e.g. built in your laptop / phone) or a headset.
• Acceptance of the Meta Data collection consent form (to be signed if you are interested).
• Timeline and volume:
• The project has already started and is expected to last until the end of June approx.
• 1-2 hours per day will need to be spent in order to deliver 2 hours of total valid audio hours at the end. 10 hours will be spent in total with recordings.
• Once the training and phases are completed, you will need to record your portion within the given timeframe.

• Tool: Squirrel (web-based application). Instructions will be provided by our client’s production team.
• Rate and payment: 15€/h.
• Each recording submitted will be reviewed to verify compliance with the provided instructions. Only recordings that successfully pass the review will be eligible for compensation. The final earnings will be based on the total number of valid recordings (PLEASE NOTE THAT WE ARE CHECKING WITH OUR CLIENT ON WHAT BASIS THEY WOULD CONSIDER THAT A RECORDING DOES NOT MEET THE APPROPRIATE REQUIREMENTS).
For every hour of valid recorded audio, you will be paid for 4 hours of work. Here you have more information regarding the validation process of the clips:
The reasons for invalidation are mainly these three:
Background noise – background noise in the clips or disturbing white noise in recordings. Speakers should record in the quite environment.
Audio does not match utterance – when the speaker record words different from the string. For example in English. “He is not bad” cant be recorded as “He isn’t bad” even if the meaning is the same.
Wrong emotional style – the speaker does not express the emotion at a sufficient level(most common one).
• Deadline: TBC

Let me know if you would be interested as soon as possible. 😊