FreeTalk User Study

The purpose of this user study is to incorporate human feedback into the evaluation of our results. It takes about 10–15 minutes in total.

The study is divided into two parts. In the first part, you will participate in an A/B comparison, where you will be asked to choose which animation (left or right) demonstrates better lip movement and which animation more effectively conveys a given emotion.

In the second part, you will be shown a set of qualitative results. For each example, you will be asked to rate both the quality of the animation and the quality of the expressed emotion on a scale from 1 to 5.

Please do not focus on the quality of the mesh or visual appearance. Instead, base your evaluation solely on the quality of the animation and the conveyed emotion.

Your answers are anonymous — no personal information is collected, only your choices and a random session id.

Part 1 — A/B comparison

In this part of the study, you will be presented with pairs of animations (left and right). For each pair, you will be asked to select: which animation demonstrates better lip synchronization with the audio, and which animation more effectively conveys the given emotion.

Please focus only on the quality of the animation and the conveyed emotion. Do not consider the mesh quality or overall visual appearance when making your decision.

Part 1 — Comparison of

Which animation is better?

Intended emotion for this clip:

Part 2 — Rating

Use the following 5-point scale to evaluate each animation:

Lip-sync quality

• 1 – Very Poor: Lip movements are clearly out of sync and distracting.
• 2 – Poor: Noticeable mismatches that make the speech harder to follow.
• 3 – Fair: Generally synchronized, with occasional minor mismatches.
• 4 – Good: Accurate and natural, with only subtle, barely noticeable issues.
• 5 – Excellent: Perfectly synchronized, highly natural, and fully believable.

Emotion quality

• 1 – Very Poor: The intended emotion is not recognizable or is misleading.
• 2 – Poor: The emotion is weak or inconsistent, making it difficult to identify.
• 3 – Fair: The emotion is somewhat clear, but not strongly or consistently expressed.
• 4 – Good: The emotion is clearly conveyed, with only minor inconsistencies.
• 5 – Excellent: The emotion is expressed clearly, naturally, and convincingly.

You may watch each video as many times as needed before assigning a rating. Focus only on how well the lip movements match the audio and how clearly the intended emotion is conveyed. Do not consider other aspects such as mesh quality or visual appearance.

Each face shape below was never seen during training.

Part 2 — Animation of

Rate this animation

Submitting your answers…

Please wait a moment.

Thank you!

Your responses have been recorded. We really appreciate your time.

Something went wrong

Your answers could not be submitted automatically. Please copy the text below and send it to federico.04021997@gmail.com.