Remote AI Output Evaluator — Flexible Hours

Outlier; San Antonio (TX); Remote; USD 14,000 - 21,000; Full time

Job summary

Outlier is seeking detail-oriented English speakers to review AI-generated media on a flexible remote schedule. You’ll hold live conversations with AI voice models, rank naturalness, and explain why outputs look or sound off, helping improve real-world AI performance.

Responsibilities include comparing AI-generated images and videos, inspecting outputs for issues like distortions or misalignment, and rating across adherence, quality and realism with concise explanations in English.

Qualifications

• Fluent English speaker with strong written communication and comfort with natural, everyday conversational speech.
• Strong attention to detail and patience to listen closely and replay or zoom in when something seems off.
• Nice to have: Pursuing or holding a Bachelor's degree or above.

Responsibilities

• Holding a live, scenario-based conversation in English with AI voice models, then ranking how natural, useful, and clear each one sounded.
• Comparing pairs of AI-generated images and videos and judging which one better follows the prompt.
• Inspecting outputs closely, including zooming in and replaying, to spot problems like distorted anatomy, flickering or unnatural motion, text errors, and audio that doesn't match the video.
• Rating each output across dimensions like adherence, quality, and realism, then writing a specific explanation in English that points to what you saw or heard.

Skills

Fluent English

Attention to detail

Clear written communication

Decision making

Education

Bachelor's degree or above

Job description

Outlier is seeking detail-oriented English speakers to review AI-generated media on a flexible remote schedule. You’ll hold live conversations with AI voice models, rank naturalness, and explain why outputs look or sound off, helping improve real-world AI performance.

Responsibilities include comparing AI-generated images and videos, inspecting outputs for issues like distortions or misalignment, and rating across adherence, quality and realism with concise explanations in English.