Kling Avatar Pro is a mouth pass

Run Kling Avatar Pro when the finished cut shows a mouth that has to match a line you already recorded. Skip it when nobody on screen is speaking, when you still need to invent the room, or when the performance you care about lives in a driving video instead of a wav.

That is a when-to-use rule, not a recitation of the meter. The catalog card is already written: professional talking avatar, lipsync content type, audio-to-lipsync and image-to-lipsync, image required, 4K ceiling, 58 credits as the listed floor of a per-second bill. This page is the routing test you run before you spend that floor.

When the mouth is the job

The cut is a close-up or a head-and-shoulders of one person. Viewers will watch the jaw. The line is locked — a founder VO, a cloned read, a TTS take you already approved. The still is locked — wardrobe, crop, lighting, the face you will recognise next week. Only then is Avatar Pro the right row.

Use it for:

• A presenter who has to say a specific script on camera, not “talking energy” under b-roll.
• A UGC-shaped take where the product is already in the plate and the claim is already in the audio file.
• A 4K talking close-up for a landing page or an ad that will be paused on the face.
• A localisation pass where the picture is the same plate and the new language is a new track. The mouth pass follows the new wav. It does not invent Spanish from an English prompt.

The door is AI lipsync. The still comes from an image row or a photograph. The track comes from text to speech or a recording. Avatar Pro is the join. If you cannot name those two files, you are early.

Length follows the track. There is no 5s / 10s chip on this row because the model is not a clip generator with a duration menu. A twelve-second line is a twelve-second talking picture. Cut the wav to the sentence you will ship. Throat-clearing is billed.

4K is the output ceiling, not a rescue. A soft phone still becomes a 4K view of softness with a working mouth. Bring a plate that can hold the resolution you intend to deliver.

When you should skip the row

Skip Avatar Pro when the mouth is not on camera. A product orbit, a screen recording, a kitchen demo shot from above, a slideshow of stills with VO — those need a voice file laid under picture, not a 4K lipsync. Paying the 58-credit floor to animate a jaw nobody will see is the expensive way to attach audio.

Skip it when the scene does not exist yet. Background, wardrobe, the product in hand, the likeness: if they are missing from the plate, Avatar Pro will not design them. “Talking host in a bright studio, holding the bottle” is a stills brief. Generate or shoot the still. Then come back.

Skip it when the performance you want is a face acting a clip, not a mouth matching a track. That is motion control: a driving video of expressions and head movement mapped onto a reference still. DreamActor V2 is that job. Avatar Pro will not watch a driving take and copy the eyebrow. It will listen to a wav and move the mouth.

Skip it when you only needed a scratch to hear the line. TTS is the scratch. Avatar Pro is the picture of the scratch. Do not 4K-lipsync a line you will rewrite after lunch.

Skip it when the crop is a wide establishing shot. A talking-avatar row is a face row. A wide plate with a tiny head will still spend the floor. Crop for the jaw, or pick a different generate.

The ranked talking-head shelf is a different question from this one. This page does not rank Avatar Pro against every lipsync SKU. It asks whether this cut is a mouth cut.

Picture first, track second, 4K last

Order is the product. Reverse it and you either lipsync the wrong face or 4K-render a throwaway line.

1. Lock the still. Front-facing enough that the mouth is visible. Light that matches the eventual delivery, not a beauty crop that hides the jaw. If a product has to appear, it is already in the frame. If a label has to read, letter it on an image row and proof it before this pass. Motion will not typeset.

2. Lock the track. One speaker. The words you will actually publish. No bed you plan to replace, no temp joke. Clone or TTS on a speech row; this SKU does not write the voice. The 58-credit figure is the lipsync pass, not the wav.

3. Run Avatar Pro. Image in, audio in, talking picture out. Review the mouth, not the room. If the room is wrong, you are back at step 1. If the line is wrong, you are back at step 2. Prompting the lipsync model to “make the office warmer” is not a lever it has.

4. Caption after the take exists. A moving mouth is not a transcript. Add captions transcribes what was said. Do not bake a headline into the still and call it a caption. Do not type the line into the Avatar Pro prompt and hope it typesets on screen.

If the week is “we need a talking ad by Friday,” this order still holds. Parallelising stills and speech is fine. Parallelising Avatar Pro with an unapproved still is how you buy a 4K miss.

Article 50 still applies if the talking picture resembles a real person and you publish it to an EU audience as authentic-looking media. Label it. The template is Article 50 disclosure. A lipsync of a generated face is not exempt because the mouth is good.

What Avatar Pro will not fix

It will not invent resolution the plate never had.

It will not hold a second character. One still, one track, one mouth.

It will not replace a brand kit. Wardrobe and grade are in the pixels you uploaded.

It will not dub. If you need Spanish, make a Spanish track, then run the mouth pass on that track. Avatar Pro is not dub a video.

It will not be your scene generator because the Kling roster also contains text-to-video. Those rows invent shots. This row performs on a picture. Mixing them in one prompt is how a talking close-up becomes a new person in a new room saying a line you already recorded for someone else.

The meter post — credits, 4K, categories — is Kling Avatar Pro: the mouth pass after the picture exists. Read that when you need the sticker. This page is the gate: is the mouth on camera, and do both files already exist?

If yes, sign up and run Avatar Pro as the last picture job, not the first.

FAQ

Do I use Avatar Pro to design the host?

No. It requires an image. Design or photograph the host on a stills row, or use a real plate. Avatar Pro is the talking pass on that plate.

The cut is VO over product footage. Do I still run it?

No. Lay the voice file under the picture. Avatar Pro is for when we see the mouth. A closed-mouth product turn with a 4K lipsync bill is the wrong object.

Can I skip TTS and type the line into Avatar Pro?

No. Audio-to-lipsync means a track. The model hears a file. A sentence in the prompt is not the performance.