Use Image-to-Video to animate an existing image. If the only goal is a face speaking your recording, Talking Photo is the more direct starting point.
Get the right result: check these first
Start from the right input. Attach the image you want to animate. Use Text-to-Video if you only have a written description; use Talking Photo if the goal is a portrait speaking your recording.
Audio or end-frame error. Uploaded audio and an end frame cannot be combined in the current audio-input workflow. Remove the conflicting input and follow the model-specific error; buying credits does not remove this restriction.
Text or details change. Inspect any lettering, logo, face, or detail that must stay unchanged throughout the clip. An instruction to preserve it is not a guarantee. Keep the source and result when requesting a review.
Still stuck? Contact support with the tool name, project link or visible ID if available, model, selected duration, audio/end-frame/reference inputs, and the timestamp of the unwanted change. Keep the project while it is reviewed. How to send useful troubleshooting details.
Animate your image
Open Image-to-Video and sign in.
Upload the starting image. Use a clear source with the subject visible.
Describe the motion you want, such as “The subject smiles and turns slightly; keep the camera still.”
Choose a model, supported duration, and resolution. Review the aspect ratio or Auto shape shown for your inputs.
Review any required fields and the credit estimate, then generate.
Check the completed result in History or My Library and download it.
Add your own audio
Select LTX-2.5, keep Audio enabled, and use Drop audio or browse. Add Audio offers upload, Text to Speech, voice cloning, and presets. You can use a starting image with your recording. See the audio workflow.
Why did the end image or references become unavailable?
Inputs are model-specific. Uploaded audio currently cannot be combined with an end image or additional reference/Character attachments in this LTX-2.5 workflow. Remove the conflicting input, or choose which feature is most important for this clip. A paid plan does not override these model restrictions.
On supported reference workflows, adding a Character or reference can cause the starting image to be generated from those references. If you see “Auto-generated from characters,” the tool is not preserving a separately chosen start image in that configuration. Remove the references if the exact starting image is essential, then inspect the controls again.
Do the thumbnails beside Add References mean references are attached?
No. Those small thumbnails are quick-add shortcuts from your saved library, not a count of references attached to this generation. Clicking one adds its @name chip inside the prompt. Typing an @name that matches a saved Character can also select that Character.
Before generating, check the prompt for reference chips or @name mentions, including anything restored from an earlier draft or creation. To stop using a reference for this generation, remove its chip or mention from the prompt; you do not need to delete it from your saved library. Start and end images are separate inputs—check those too. If the prompt and the visible inputs still do not match what you intended, send support a screenshot before spending credits on another attempt.
Why is aspect ratio set to Auto, or 4K unavailable?
A submitted start or end frame can determine the video shape, so aspect ratio may show Auto instead of an editable choice. Prepare the source image in the intended shape and review the visible crop or preview before generating. A ratio typed in the prompt does not override the submitted frame. Where the aspect-ratio control is available, choose one of the options actually offered for that model.
4K requires a compatible model and an eligible paid plan. Kling 3.0 offers 4K for supported inputs; LTX-2.5 does not. Review the current model, input restrictions, resolution choices, and any plan lock before generating. Buying a credit pack does not by itself unlock paid-plan resolution settings.
Make longer videos and consistent scenes
Use the duration options shown for your model. For a longer story, create individual clips and assemble them afterward. Saved Characters can help with consistency in compatible modes; they do not guarantee identical faces or exact motion. Avoid changing several settings at once when refining a result.
Upload or generation problems
Read the exact validation message and confirm the upload completed. If a supported file repeatedly fails, or a completed output is blank, cut short, or missing expected audio, stop repeating the request. Send support the route, model, input type, selected duration, error, and project link if available. Check generation status and credits.