How to pick a lipsync catalog row

Pick a lipsync row from the files you already have, not from a brand name. Audio-to and image-to need a still plus a voice. Video-to needs a talking clip plus new audio. Kling Avatar Pro is 58 credits at 4K. Kling Lipsync is 7 credits, video-to, 4K. VEED Fabric 1.0 is 40 credits at 720p. VEED Lipsync is 4 credits, video-to. Match the inputs. Do not buy 4K avatar when you already have a video that only needs a new mouth.

This is a row picker. It is not a lipsync quality essay.

Inventory the inputs before you open a row

Write down what exists. The catalog will not invent a missing file.

Audio — a voice take: VO, TTS, a dub, a translated track. Duration of the audio is usually duration of the mouth.

Still — a face (or a character) you would approve as a freeze-frame. Front-on enough to see the mouth. Not a pack shot. Not a 16:9 crop of a group.

Video — a clip where a mouth already exists: a camera talking head, a generated talking head, an avatar you already made. The body and the room are in the pixels. You only want the lips to follow a new track.

Sort into one of three jobs:

Job	You have	You do not have	Family
Audio-to / image-to	Still + audio (or still + script on rows that TTS for you)	A talking video	Avatar / fabric
Video-to	Video + new audio	Need to keep the existing picture	Mouth pass
Neither	A prompt	Any of the above	Not lipsync. That is T2V.

If you have only a still and no audio, you do not have a lipsync job yet. Run TTS first (Gemini 3.1 Flash TTS is 4 credits, inline audio tags) or record. If you have only audio and no face, you still need a still. If you have only a prompt that says “talking influencer,” you want a video generate, then maybe lipsync later — or a native-audio row if the line does not have to be exact.

Lipsync video is the editing-job door. The row under it still has to match this inventory.

Rights: you need the right to use the face and the voice. A celebrity still plus a cloned voice is not a catalog problem. It is a legal one. Stop.

Audio-to and image-to: you are building a talker

These rows take a still (required on the cards below) and drive the mouth from audio or from a script.

Row	Categories on the card	Resolution listed	Credits
Kling Avatar Pro	audio-to-lipsync, image-to-lipsync	4K	58
VEED Fabric 1.0	audio-to-lipsync, image-to-lipsync	720p	40
VEED Fabric 1.0 Text	text-to-lipsync, image-to-lipsync	720p	40

Pick Kling Avatar Pro (58cr, 4K) when:

The still is approved.
You have a voice track (or will attach one).
Delivery wants 4K, or the face is a close-up that will be cropped.
You are willing to pay 58 credits for that identity.

Pick VEED Fabric 1.0 (40cr, 720p) when:

Same still + audio job.
720p is enough (a 9:16 feed, a draft, a mid-shot).
You do not need the 4K avatar card.

Pick VEED Fabric 1.0 Text (40cr, 720p) when:

You have a still and a script, not a voice file. The card’s job is: turn a photo into a talking video from a script; the voice is auto-generated to match the character.
You accept that you are not bringing a brand-read VO. If the line must be the founder’s voice, this is the wrong row — TTS or a recording, then audio-to.

Procedure:

Lock the still. Mouth visible. Lighting you can live with. Crop to the ratio you will ship.
Lock the audio (or the script on the Text row). Length is length.
Attach both. Do not type choreography into the prompt and skip the still.
Generate. Watch the mouth at 1×, then scrub plosives (p, b, m).
If the identity is wrong, the still was wrong. Do not switch to video-to to “fix a face” you never approved.
Caption after the mouth is right. Transcribe the audio you actually attached.

Do not pick Avatar Pro because the name says Pro when you already have a talking video. That is the next section, and it is 7 credits or 4, not 58.

Video-to: you already have the picture

These rows take an existing video and a new audio track. The body stays. The mouth changes.

Row	Category	Resolution listed	Credits
Kling Lipsync	video-to-lipsync	4K	7
VEED Lipsync	video-to-lipsync	not listed	4

Pick Kling Lipsync (7cr, 4K) when you have a 4K (or good enough) talking clip and a new track: a translation, a corrected line, a different read. This is the cheap mouth pass on a file that already exists.

Pick VEED Lipsync (4cr) when the job is the same and you want the cheapest video+audio lipsync on the catalog. The description is explicit: simple video + audio input.

Procedure:

Confirm the source video is the picture you will keep. If the face is wrong, video-to will faithfully lip-sync the wrong face.
Confirm the new audio is the line you will keep. Duration should be close to the source. A 20-second VO on a 6-second clip is not a mouth pass; it is a rebuild.
Attach video + audio to the video-to row.
Generate. Check teeth, jaw, and whether the rest of the performance still matches (a smile that disagrees with an angry line).
If you also need a dub of an existing language, dub video is the language job; lipsync is the mouth job. Sometimes you run both. Do not expect one row to translate and re-perform the body.

LTX 2.3 audio-to-video lipsync exists on the catalog at 20 credits, 1080p, as a mouth pass. Use it when that card matches the file. Do not invent it as a substitute for the four rows in the title.

Native-audio T2V is not lipsync. Veo and Seedance will speak in the same pass as the picture. You do not control the face from a still, and you do not attach a brand VO. If the line must be exact, silent plate + TTS + video-to lipsync (if you have a mouth to drive) or image-to lipsync (if you have a still) is the controlled path.

Refuse the mismatches

Walk this list before you spend:

Still + “make them dance and talk.” Dance is motion control (still + driving clip). Talk is lipsync (still + audio, or video + audio). Two rows, two files. A lipsync row will not choreograph a dance from a sentence.
Video you like + no new audio. There is nothing to sync. You are done, or you are captioning.
Audio + no face. Make or shoot the still first.
4K delivery + Fabric 1.0. Fabric lists 720p. Avatar Pro lists 4K at 58 credits. Pick on purpose.
Existing 4K talker + Avatar Pro. You are about to pay 58 to rebuild a picture you already have. Kling Lipsync is 7.
Script only, no still. Fabric Text still wants the photo (required on the card). T2V with native audio is the prompt-only door, with all the identity drift that implies.

After the mouth is right, burn captions from the audio you actually shipped. Do not transcribe an old take you replaced.

If you need a studio seat to attach the still or the video and run the matching row, sign up. Bring the files. Do not bring only the word “avatar.”

FAQ

Kling Avatar Pro or Kling Lipsync?

Avatar Pro (58cr, 4K) is audio-to / image-to: you have a still. Kling Lipsync (7cr, 4K) is video-to: you already have a talking clip. Same family name, different inputs.

Why would I pay 40 credits for VEED Fabric instead of 4 for VEED Lipsync?

Fabric builds a talker from a still. VEED Lipsync retimes a mouth on a video you already have. If you have the video, 4 credits is the row. If you only have a photo, 4-credit video-to cannot start.

Can I lipsync from a text prompt with no still and no video?

Not on these cards. Image-to rows require a still. Video-to rows require a video. A prompt-only talking clip is a native-audio generate, not lipsync.