Videos and Character Studio
Make product videos and presenter (UGC) videos, build reusable AI presenters with looks and voices, and know which provider each step needs.
6 min readUpdated Oct 17, 2026Applies to Adcraft 1.0.0
On this page
Adcraft makes two kinds of video:
- Product video: about 15 seconds of product scenes with captions, a hook and a brand end card.
- Presenter (UGC-style) video: about 30 seconds with an AI presenter speaking a script, product B-roll and captions.
Both are rendered in three sizes, 9:16 (1080 × 1920), 1:1 (1080 × 1080) and 16:9 (1920 × 1080). You choose the primary size, and every video is also resized to the other two.
Product video
Choose Make a video on a selected video concept, or start from Videos.
How it is made
- Storyboard. The concept's script is split into up to eight scenes of 2 to 5 seconds each, plus a 2.5-second end card with the headline and call to action.
- Scene stills. The default image model makes a still for each scene, using the product cutout as a reference.
- Image to video. The chosen video model animates each still. Clip lengths snap to what the model supports.
- Assemble sizes. Remotion renders the three sizes with captions, the hook over the first scene and the end card with the logo.
Progress for each step shows on the video page and in the activity tray. Owners and editors get an email when it finishes, if Resend is configured.
Video models
| Model | Clip lengths | Notes |
|---|---|---|
| Kling 3.0 | 5 or 10 s | The default |
| Veo 3.1 | 4, 6 or 8 s | 16:9 and 9:16 only |
| Seedance 2.5 | 5 or 10 s | |
| Seedance 2.0 | 5 or 10 s | Low-cost draft option |
All four run on fal.ai. Admins can add more or enable the Replicate and Runway examples; see AI models and providers.
Edit and regenerate
On the video page each scene has a Caption (and, on the first scene, Hook text), which re-assemble the video for free, and a Scene prompt with Regenerate for that shot only. Re-assemble rebuilds the sizes from existing clips; Regenerate everything starts again.
A full run costs 20 credits, whichever video model you pick. Regenerating some scenes costs a share of that; re-assembling is free.
Important
Product-video rendering uses Remotion, which has its own licence: free for individuals and organisations of up to three people, paid for larger ones. Read THIRD-PARTY-NOTICES.md and Remotion's licence before commercial use.
Presenter (UGC) video
There are two ways in:
| From a concept (Make the UGC video) | From Character Studio (Create an ad) | |
|---|---|---|
| Presenter | A presenter from HeyGen's library | One of your saved characters, in any of its looks |
| Voice | A voice from the catalogue | The character's voice, or any catalogue or custom voice |
| Script | The concept's script | One of six story templates, with three suggested openings, edited by you (up to 75 words) |
| Without keys | Runs with placeholders | Needs fal.ai and HeyGen before it can start |
| Credits | 40, charged when it finishes | 40, reserved at the start and refunded if it fails |
How it is made
- Voice-over with ElevenLabs (
eleven_multilingual_v2), unless a HeyGen voice is chosen. - Presenter rendered by HeyGen from the library look, the character's HeyGen look or its portrait.
- B-roll: two short product clips from the chosen image and video models.
- Assembly: the presenter is the base track, B-roll cuts in, captions follow the script, and the hook sits in the lower third.
Every presenter video carries a visible AI-generated label in the corner.
Story templates
| Template | Structure |
|---|---|
| Meet your next everyday essential | Introduction, product, invitation |
| Let the details do the talking | Curiosity, close-ups, invitation |
| Start with a question | Question, answer, invitation |
| Make it easy to understand | Context, explanation, next step |
| Introduce something new | Announcement, introduction, invitation |
| A closer look | Discovery, detail, explore |
Each template has an opening, a product message and a call to action, with Explore three openings for alternative hooks. On-camera energy (Energetic, Natural or Calm) and optional Gesture notes shape the delivery; how much they change depends on the kind of presenter.
Character Studio
Characters is a cast of reusable presenters for a brand. The sidebar opens Create an ad, My characters, Presenter library, Voice library and Design a look. It needs the plan's UGC feature switch to be on.
New character
| Source | How | Needs | Credits |
|---|---|---|---|
| Generate a face | A fictional adult portrait from your description (at least 15 characters) | A connected image model | The image model's credits |
| From a photo, or Describe them for HeyGen | A HeyGen photo avatar or prompt avatar | HeyGen | 5 |
| Clone a real person | A HeyGen digital twin from footage, with the voice cloned from the same footage | HeyGen | 25 |
Digital twin footage must be an MP4 or MOV of 15 seconds to 10 minutes, up to 600 MB, with the face in frame throughout. You must confirm the person has agreed to be cloned; HeyGen then needs them to record a short consent statement. The character card shows the consent status with Copy consent link and New consent link (each link is valid for 24 hours), and the twin can be used once consent is approved.
Warning
Only clone people who have given you permission, and follow HeyGen's terms. Some consent options are only available on HeyGen enterprise plans.
Looks
A look puts a character in a new outfit or setting.
- Describe it: your words, made with an image model, keeping the face from the portrait.
- HeyGen look packs (3 credits per look): Modern Corporate, Stylish Business-Casual, Scandinavian Minimal, Cool Tech / Slate, Healthcare, Keynote Power Presenter, Fitness and Automotive Dealership make five looks each; Business Casual and Podcast make two.
- Remix a library look: start from any look in HeyGen's presenter library (two looks).
- Your own words on HeyGen (one look).
You can choose a women's or men's wardrobe and a portrait or landscape frame. Looks that fail are refunded.
Presenter library
With HEYGEN_API_KEY set, Adcraft indexes HeyGen's public presenter library in the background (at the time of writing, about 1,400 people and 25,000 looks). Search and filter it, then use a look in an ad or remix it for your own character. The size and content of the library are HeyGen's and can change.
Voice library
The voice library combines HeyGen's public voices and your ElevenLabs voices (or five stock ElevenLabs voices if you have no ElevenLabs key). You can audition voices before choosing one.
| Tool | Provider | Notes |
|---|---|---|
| Clone your voice | HeyGen instant clone | One recording (MP3, WAV or M4A, up to 10 MB), with a permission checkbox. No Adcraft credits. |
| Design a voice | HeyGen voice design | Describe the voice (at least 20 characters), compare up to three results and keep one. The others are deleted. |
Custom voices stay private to the brand that made them and use voice slots on your HeyGen account. Each brand can hold 2 custom voices by default (the HeyGen voices per workspace guardrail). Voice cloning through ElevenLabs is not offered.
Without keys
| Missing | What happens |
|---|---|
FAL_KEY |
Product video: placeholder stills with a slow zoom, assembled into a real file. 20 credits are still charged. |
ELEVENLABS_API_KEY |
Presenter video from a concept: a silent voice track (HeyGen voices still work) |
HEYGEN_API_KEY |
Presenter video from a concept with no presenter chosen: a "presenter card" placeholder labelled "AI PRESENTER · PREVIEW", and 40 credits are still charged. With a library presenter chosen, the run fails with "HeyGen is not connected". Character Studio: HeyGen options are disabled. |
Not in this version
- Product-video stills always use the default image model, not a model you pick for the video.
- Caption style and video fonts are fixed. Videos use the bundled DM Sans and Instrument Serif rather than the brand kit's fonts.
- Some labels are out of date. The video page always describes the voice as ElevenLabs and the presenter as a "licensed stock avatar", and the concept form says custom faces and cloned voices need "a consent flow first", even though Character Studio now provides one.
- Rendering happens on your server. There is no cloud rendering option; size your server for it (see Requirements).