Here is the short version. VEED Fabric 1.0 turns a single image plus an audio track into a lip synced talking video. You can run it today on Fal.ai. Pricing is simple. 480p costs $0.08 per second. 720p costs $0.15 per second. Clips are currently capped at 30 seconds per generation, and you can stitch clips for longer outputs.
\n
What Fabric 1.0 actually does
\n
Fabric 1.0 is an image-to-video API built to map facial motion to audio and render a talking head video from a static image. You feed it a face or character image and a voice track. It outputs a short video with realistic mouth motion in sync with the audio. That is it. No camera, no stage, no scheduling. For a lot of marketing, training, and product explainers, that is the job.
\n
Supported inputs
\n
- \n
- Images: jpg, jpeg, png, webp, gif, avif
- Audio: mp3, ogg, wav, m4a, aac
- Optional: AI text to speech if you do not have a recorded voice
\n
\n
\n
\n
Output
\n
- \n
- Talking head video with lip sync aligned to the supplied audio
- Resolution options: 480p or 720p
- Max length per generation: 30 seconds
\n
\n
\n
\n
Pricing that is easy to forecast
\n
Most image to video tools hide cost behind odd credit systems. Fal.ai lists Fabric 1.0 at a straight per second rate. This matters when you are budgeting weekly content or building a product that calls the API on demand.
\n
- \n
- 480p: $0.08 per second
- 720p: $0.15 per second
\n
\n
\n
Cost math is straightforward. 10 seconds at 480p is $0.80, 30 seconds is $2.40. 10 seconds at 720p is $1.50, 30 seconds is $4.50.
\n
\n
That clarity is useful if you plan to generate content at scale or gate output by plan limits in your own app. Inside VEED’s paid plans you can use credits as well. Each Fabric generation currently consumes 240 credits. The platform also allows stitching multiple 30 second renders to produce longer videos.
\n
Who this solves a problem for
\n
- \n
- Marketing teams with product explainers, teasers, and social ads that need a face and voice without booking a shoot
- Creators and podcasters who want a talking avatar or character for clips
- Support and training teams producing update summaries or micro lessons
- Education workflows where a virtual instructor reads short scripts
\n
\n
\n
\n
\n
It is also a fit for any app that needs a quick talking head without shipping video assets or managing live capture. The API route through Fal.ai makes that integration direct.
\n
Image and voice options
\n
Fabric accepts your own images or AI generated portraits. You can also pick a character from a library if you are working inside VEED. For audio, record your own, upload a track, or generate text to speech with AI voices. The lip sync is as good as the clarity of the audio you give it. If the input has noise, reverb, or inconsistent levels, expect the mouth motion to feel less precise.
\n
Quality notes
\n
- \n
- Fabric targets studio grade talking head output. It is designed for production use, not only for demos.
- It handles a range of styles. Realistic portraits, illustrated characters, and animated images all work, though a clear frontal face and consistent lighting help.
- Mouth motion follows phonemes from the audio. Timing quality rises with clean, well paced speech and neutral background noise.
\n
\n
\n
\n
Limits to plan around
\n
- \n
- 30 seconds max per render today
- 480p and 720p are priced and available. If you need higher resolutions, plan to upscale after the fact or watch for future tiers
- Strongest results come from a straight on face with visible lips and minimal occlusion. Side profiles or heavy stylization can reduce sync quality
\n
\n
\n
\n
How to run it through Fal.ai
\n
Fal.ai exposes Fabric 1.0 as a hosted model with an API. You send the image and audio to the model endpoint and receive a video output. Start here: Fal.ai Fabric 1.0 model page. The page shows accepted file types, endpoint details, and returns. If you prefer a low touch workflow, VEED offers Fabric inside their editor with credits and project tools like captions, storage, and sharing.
\n
A practical setup flow
\n
- \n
- Prepare your input image. Aim for 1024 on the long side or similar, clear face, straight on, good contrast
- Prepare your audio. Use 48 kHz WAV or a high bitrate MP3. Keep it dry and clean. Target 20 to 30 seconds per clip
- Call the Fal.ai endpoint for Fabric 1.0 with the image and audio. If you want 720p, set the output parameter accordingly
- Collect the result and store it in your pipeline along with metadata for tracking cost by length
- For longer scripts, split into segments under 30 seconds, render multiple clips, then stitch
\n
\n
\n
\n
\n
\n
Workflow ideas
\n
- \n
- UGC style ads. Draft the script, generate voice with AI, render a talking avatar, add captions and brand frames
- Product release notes. Monthly 20 second summaries from a branded character
- Education micro lessons. 30 second concepts with a virtual teacher
- Podcast clips. Cover art that talks, synced to a short quote
\n
\n
\n
\n
\n
Integration tips if you are building on top
\n
- \n
- Gate user requests by duration. Cost scales linearly with seconds. A 25 second 720p clip is $3.75. That is predictable for pricing your own tiers
- Pre flight checks. Validate image aspect ratio, face position, and audio length before the API call to reduce fails
- Batch jobs. Queue clips and run during off peak hours if your schedule allows, then assemble longer outputs
- Fallback plan. If the input face is weak, switch to a library character with known quality
\n
\n
\n
\n
\n
Positioning and how it compares
\n
VEED calls Fabric the first AI talking video model. The core point for buyers is simpler. It is fast to set up, the outputs look good, and the cost is clear. If you only need talking head clips with reliable lip sync, Fabric is a direct answer. If you need cinematic scene changes or full scene composition, that is a different class of tool.
\n
If you are gathering options for a broader video stack, you might also want to look at focused editors and model updates covered here. For text guided video editing, see my notes on Decart Lucy Edit: Decart Lucy Edit. For model based video reasoning improvements on the creative side, here is a look at Ray3 inside Firefly: Ray3 Lands In Adobe Firefly. For audio generation on Fal.ai that pairs well with talking heads, I covered cost and quality on MiniMax Music 1.5: MiniMax Music 1.5.
\n
Where VEED fits in the picture
\n
VEED offers Fabric inside its editor with storage, captioning, and sharing. Paid plans include AI credits, and each Fabric generation uses 240 credits. If you prefer a no code approach with project management features, that route is straightforward. VEED has a Fabric API in development with a waitlist. If you need API access now, use Fal.ai.
\n
Edge cases and content policy risk
\n
- \n
- Faces work best front facing with visible mouth. Obstructions like hands or masks reduce quality
- Cartoonish styles can be fine if the mouth is distinct and the head is stable in the frame
- For brand compliance, pre approve character images and voices. Keep a record of licenses for any stock or generated portraits you use
- Do not impersonate real people without consent. Build an approval step into your workflow if you accept user uploads
\n
\n
\n
\n
\n
Why this is useful right now
\n
Most teams do not need a giant stack of animation tools for simple talking segments. They need a believable face, timed to a script, that ships today. Fabric does that with good consistency, and at a price you can quote to a boss without a long explanation. The 30 second cap is a limit, but stitching is fine for social and internal training. If you need continuous two minute talking head monologues, you will be doing segment workflow anyway to keep pacing tight.
\n
Practical checklist
\n
- \n
- Decide on 480p or 720p before writing scripts, so you know cost per deliverable
- Keep scripts under 28 seconds to leave headroom for intro and outro frames
- Record clean audio in a quiet room, speak at a steady pace, and normalize levels
- Use consistent, frontal portraits with neutral backgrounds
- Batch test three images per voice to pick the most natural lip motion, then lock that combo for your brand
\n
\n
\n
\n
\n
\n
Budget planner quick math
\n
- \n
- Ten 20 second clips at 480p: 10 × 20 × $0.08 = $16.00
- Ten 20 second clips at 720p: 10 × 20 × $0.15 = $30.00
- Fifty 30 second clips at 480p: 50 × 30 × $0.08 = $120.00
- Fifty 30 second clips at 720p: 50 × 30 × $0.15 = $225.00
\n
\n
\n
\n
\n
Credit based users inside VEED can map these to monthly allowances. If you are productizing the API, that same math makes it simple to set per minute or per second pricing in your own tiers.
\n
Common failure modes and how to avoid them
\n
- \n
- Low resolution or blurry faces. Fix: source sharper portraits, avoid heavy compression
- Side angles or occluded mouths. Fix: keep the face straight on and unobstructed
- Noisy or reverberant audio. Fix: record closer to the mic, reduce room echo, and apply light noise reduction
- Overlong scripts. Fix: keep to 20–28 seconds per segment and stitch
\n
\n
\n
\n
\n
API integration notes
\n
Fal.ai provides a clean endpoint for programmatic use. Send the image and audio, choose 480p or 720p, and fetch the video result. If you want to centralize editing, captioning, and sharing, you can also run Fabric inside VEED and manage assets there. Start with the Fal.ai model page: Fal.ai Fabric 1.0.
\n
Where to try it
\n
Run Fabric 1.0 on Fal.ai here: Fal.ai Fabric 1.0. If you prefer an editor first flow with credits and built in captioning, open VEED and look for Fabric inside the AI features.
\n
Bottom line
\n
Fabric 1.0 is a direct tool for a clear job. Image plus audio in, talking video out. The price per second is transparent, the format support is sensible, and the 30 second render limit is workable for most short form content. If you are building a repeatable talking head pipeline, this is worth adding right now.
