VEED Fabric text vs audio input

VEED Fabric 1.0 Text is 32 credits, lipsync, 720p, audio true, image required, categories text-to-lipsync and image-to-lipsync. VEED Fabric 1.0 is 32 credits, lipsync, 720p, audio true, image required, categories audio-to-lipsync and image-to-lipsync. Same sticker. Same ceiling. Same still gate. The fork is the track: a script the row voices, or a file you already recorded.

Both rows need a face that already exists as a still. This page is the input pick. It is also not VEED Lipsync at 3 credits. That slug is video-to-lipsync on footage. Fabric is photo-to-talking at 32.

Text description: turn a photo into a talking video from just a script — the voice is auto-generated to match the character. Audio-row description on the snapshot says professional fabric and texture video generation. The published categories are audio-to-lipsync and image-to-lipsync. Trust the categories. You bring a still and a track. You do not generate upholstery.

Duration lists are empty on both. Length follows the script or the track. Thirty-two credits is the listed sticker, not a 5s chip, and the meter runs per second after that: 7 credits at SD, 12 at HD/4K. Cut the line before you spend the floor on throat-clearing.

Text is script-in, auto-voice, 32 credits

Use Fabric Text when you have a still and words, and you do not have a voice file you must keep. The model auto-generates a voice to match the character. You are not picking a separate ElevenLabs id on this card. You are not prompting a new location. The photo is the room. The script is what the mouth says.

Write spoken sentences. A cinematic paragraph will not typeset. If the plate is wrong, stop. Generate or shoot the still on an image row. Do not ask Fabric to “make it look like a studio” in the script.

When Fabric Text is the pick:

The face is approved as a still.
The line exists as text, not as a WAV you must preserve.
720p talking is enough. There is no 4K token on these cards.
Auto-voice is acceptable to legal and to the brand.

When Fabric Text is the wrong pick:

The VO is already recorded, licensed, or cloned. That is Fabric 1.0 (audio-to-lipsync) at the same 32.
You have footage of a talking person and a new track. That is VEED Lipsync at 3 credits, or Kling Lipsync at 6 credits, 4K.
You wanted a premade library face, not your still. HeyGen Avatar V3 is 14 credits, premade-avatars, no image required.

Thirty-two credits is the catalog sticker on Text. Do not call auto-voice “free.” It is included in the 32. It is not a reason to skip the still.

Audio-in is the track you keep, same 32

Use Fabric 1.0 when the voice is the brief: a founder WAV, a cloned line, a legal-approved read. The still is still required. The mouth follows the file. If the track is wrong, the mouth will be faithfully wrong.

Same 720p. Same 32. Same empty duration list. Same audio true — here true means the row is a talking pass, and you are feeding talking audio. Do not feed silence and hope for a script. That is the Text slug.

When Fabric 1.0 (audio) is the pick:

The still is approved.
The track is approved.
You need the mouth on that photo, not on footage (footage is video-to-lipsync at 3 or 6).
720p is enough.

When Fabric 1.0 (audio) is the wrong pick:

You have no track. Use Text at 32, or record first.
You have no still. Fabric will not invent the founder. HeyGen V3 will cast a premade twin at 14.
You needed 4K lipsync on footage. Kling Lipsync is 6 credits, 4K, video-to-lipsync.

Do not run Text then Audio on the same still “to see which voice.” You will spend 64 credits to learn you should have recorded the line. If the voice must be a person, record or clone, then Audio. If the voice may be auto, Text. Pick once.

Cartesia Sonic 3.5 is 4 credits per 1,000 characters of TTS if you want a catalog voice file to feed the audio-in row. Sonic plus Fabric’s 32 is a stack you should name: Sonic for the read, Fabric for the mouth. If auto-voice on Text is good enough, skip Sonic and stay at 32.

Same 32, same 720p, different track

Production order:

Lock the still. Head and shoulders, room you approved, crop for 9:16 or 16:9 before you generate.
Decide if the voice must be a specific file.
If no: Fabric Text, script only, 32.
If yes: Fabric 1.0 audio-in, still + track, 32.
Caption the talking 720p file. Mute autoplay hides the mouth’s work.

The AI caption generator is the typesetting. Captions for silent autoplay is the plan. Burned-in captions are the short-form object. Fabric will not typeset.

UGC that is a real person still belongs here when you have a release. Fabric will not make an unreleased face safer. A premade HeyGen twin at 14 is a different likeness problem — library, not your founder.

Article 50 still applies to photoreal talking people. Thirty-two credits does not exempt a generated or animated face. Label at first exposure for EU audiences since 2 August 2026. The Article 50 disclosure template is the copy. New York’s synthetic performer rule (N.Y. GBL § 396-b, in force 9 Jun 2026) is a different object: it covers unidentifiable synthetic humans in ads, not digital twins of identifiable people. Fabric on a real founder still is the identifiable-person problem, not that statute. Do not mix them.

Sora leftovers do not become Fabric. Sora invented scenes. Fabric needs a photo. Sora’s API shuts 24 September 2026 (OpenAI Help Centre, September 2026). Map the still first.

The AI video generator will show generate tiles. Fabric is lipsync. Open the lipsync job, not a landscape prompt.

What the shared 32 is not

32 is not 4K. 32 is not video-to-lipsync at 3 credits. 32 is not HeyGen’s 14-credit premade library. 32 is VEED Fabric: photo in, talking 720p out, text-script or audio-track. Pick the input you actually have.

Run the Fabric door that matches the track in Versely.

FAQ

Is Fabric Text cheaper than Fabric 1.0?

No. Both published stickers are 32 credits at 720p. You pick by whether you have a script or a voice file.

Can I skip the still if I have a long script?

No. Both cards require an image. The photo is the scene. HeyGen Avatar V3 is the 14-credit premade path if you refuse to bring a still.

Should I use VEED Lipsync at 3 credits instead?

When you already have footage of the mouth, yes. Fabric is photo-to-talking at 32. VEED Lipsync is video-to-lipsync at 3.