A missing voice control can originate in how an element was classified, before video generation begins. Check the subject type on the actual asset. Kling’s element-library guide describes different subject categories and allows category review; character assets have voice-related choices. Do not assume every uploaded picture becomes a speaking character.
Identify what the upload actually represents
A tightly cropped jacket photograph, a mannequin wearing clothing, and a full portrait can look related to a human reviewer while representing different assets. Write one sentence stating what should be reused: the jacket design, the person’s appearance, or the whole scene. This prevents a clothing reference from acquiring an accidental identity role merely because a person appears in the picture.
Inspect the stored thumbnail, category and descriptive text together. If the card says garment and the project genuinely needs a jacket reference, the absence of a character voice option may be appropriate. If the project needs the authorized person and the classification is wrong, investigate the supported correction control. Do not change the category solely to make an unrelated button appear.
Correct the asset role before the script
Consider a hypothetical product demonstration. Image A shows a coat on a model; image B is an authorized frontal portrait of the spokesperson. The goal is to keep the spokesperson and discuss the coat. A useful asset plan separates the person from the product. Reclassifying the coat as a character would mix those roles rather than solve the missing control.
| Reference | Intended role | Classification check | Next action |
|---|---|---|---|
| Frontal portrait | Speaking person | Character identity | Review voice options |
| Coat detail | Clothing design | Garment or item | Preserve visual detail |
| Store photograph | Setting | Scene | Keep background role |
These rows describe an authored planning example. They do not imply that every mode accepts all three references or combines them successfully. Confirm the current workflow’s attachment contract separately.
Reopen the exact asset after saving
If a supported correction is made, save it and reopen the card that will be selected in the request. Check for duplicate names, an older asset still referenced by the draft, and a category change that was never saved. Record identifiers rather than relying on thumbnail resemblance. An option appearing in one card does not prove that the submitted request points to it.
When the correct category still lacks the expected control, compare the current mode, asset type and documented availability. Capture the discrepancy for support. Do not fabricate a hidden menu path or assume that another platform’s character features apply. Interface language, account availability and model version can affect what is exposed.
Define a successful fix narrowly
This investigation passes when the project uses the correctly classified asset and its supported controls are understood. It does not pass merely because a voice button appears. Actual output still needs separate identity, dialogue and sound review.
Keep the asset-role table beside an Imagild brief when those references are used there. External element-library categories are not proof of equivalent Imagild functionality. If the selected workflow accepts only an ordinary image, describe it honestly as an image reference and avoid claiming that a persistent speaking-character asset has been created.