General FAQ (Prompts, LoRAs & Generation)

Some words or phrasing may trigger the moderation system. Try replacing sensitive terms with neutral alternatives, for example:

• "girl" → "woman"

• "boy" → "man"

Also avoid explicit or borderline wording — even small changes in phrasing can help your prompt pass successfully.

The best reference is the templates inside the tool. They show realistic capabilities and expected results. If something looks difficult to reproduce — it probably is.

It depends on the type of LoRA:

• Blink LoRAs: Use any image where the face is clearly visible and well-lit.

• Standard LoRAs: Use an image that is as close as possible to the reference video (pose, angle, composition).

Use an image with one person only. This helps the model apply motion more accurately.

Yes, you can add a prompt, but it's recommended to keep it minimal. LoRAs already define motion and behavior, so additional prompts may conflict with them. In some cases, this can lead to unstable or unexpected results.

It depends on the tool:

• Image to Video: only one LoRA is supported

• Text to Image: you can use up to 3 LoRAs

Using multiple LoRAs may lead to less stable or unpredictable results, so it's recommended to combine them carefully.

LoRAs are still AI-based and may produce imperfect outputs. A certain percentage of results may include artifacts or inconsistencies — this is normal behavior. For better results: use clear inputs, follow the LoRA's intended use, and retry generation if needed.

AI does not copy 1:1 — it interprets inputs. Small differences in pose, lighting, or prompt can significantly change the result.

Consistency depends on input quality, prompt stability, and model limitations. For better consistency, reuse the same inputs and avoid changing multiple variables at once.

This usually happens when motion is too complex, the camera is unstable, or the prompt is too abstract. Use simpler scenes and stable camera references for better realism.

Yes — the best approach is iterative: adjust the prompt, change the input image, and regenerate. AI workflows often require several attempts for optimal results.

18+ LoRAs are available only for users with Trusted Status. If your account does not have this status yet, these LoRAs will not be visible.