Publish spokesperson videos without filming every one.
Script, generate, voice, translate and label AI avatar Shorts for a business, coaching offer or explainer channel, with real costs.
Write a script built for an avatar
Avatars deliver lines evenly, so the script has to carry the hook and pacing. Short sentences also reduce lip-sync errors and wasted generation credits.
- Keep Shorts under 3 minutes (YouTube's Shorts cap); for most explainers aim for 30-60 seconds, roughly 75-150 spoken words.
- Open with the viewer's problem in the first sentence, then one point per sentence.
- Spell out numbers, acronyms and brand names phonetically where the voice mispronounces them.
- Write on-screen text and B-roll notes in a second column so editing is planned before you generate.
Tool: Canva · Free option: Any text editor or Canva Docs on the free plan
Tip: Generate a 10-second test line before the full script; HeyGen and Synthesia both bill per generated minute or credit, so retakes of a full script are the main cost leak.
Choose a stock avatar or record a consented digital twin
A stock avatar is instant but generic; a digital twin looks like you but needs good footage and a recorded consent from the person being cloned.
- Stock route: Synthesia's free Basic plan lists 9 avatars and 10 video minutes; Starter lists 125+ avatars.
- HeyGen digital twin: record one uninterrupted take of at least 2 minutes (about 5 recommended), at 1080p/30fps or better, in 9:16 if you only make Shorts, with even light, a static background and only your voice audible.
- Synthesia personal avatar from video: single continuous take of 1-5 minutes, .mp4/.mov/.webm up to 2GB, one person in frame, mouth and both eyes visible.
- Record the consent clip yourself: HeyGen asks the depicted person for a short consent video; Synthesia requires consent recorded live with an on-screen passcode, by the same person, aged 18+.
Tool: HeyGen · Free option: synthesia (Basic plan stock avatars) or HeyGen Free with 1 custom avatar
Tip: You cannot clone a colleague or client by uploading their footage; each person has to record their own consent, so book them for the session rather than collecting old videos.
Set the voice
Voice quality decides whether an avatar feels watchable. A clone of your own voice keeps the brand consistent across videos and languages.
- Use the avatar tool's built-in voice first; Synthesia clones your voice as part of the personal avatar process.
- For more control, generate narration in ElevenLabs: Starter ($6/mo) includes a commercial license, about 30 minutes of TTS and Instant Voice Cloning.
- Professional Voice Cloning needs ElevenLabs Creator ($22/mo, about 121 minutes).
- Export the audio and upload it to the avatar as an audio-driven script so lips follow the recorded voice.
Tool: ElevenLabs · Free option: ElevenLabs Free (about 10 minutes of TTS, commercial license not listed) for tests only
Tip: Clone only your own voice: YouTube lists voice cloning of your own voice for voice-overs among uses that do not need the altered content disclosure.
Translate with lip-sync for extra markets
One script can reach several language audiences, but each translated version uses credits and needs a native check.
- In HeyGen, run video translation on the finished avatar video; length caps match avatar video (1 minute on Free, 30 minutes on Creator and Pro).
- Or use ElevenLabs dubbing from the same credit pool as TTS, then lip-sync in the avatar tool.
- Have a native speaker review captions and pronunciation before posting.
- Post each language on its own account or as a separate upload rather than mixing languages in one feed.
Tool: HeyGen · Free option: ElevenLabs
Tip: HeyGen does not publish a fixed credit cost per minute (it varies by model and length), so translate one 30-second short first and check how many credits it used before batching.
Edit, caption and export in 9:16
Raw avatar output is a static talking head; captions, cuts and B-roll keep viewers watching on mute.
- Export the avatar at 1080x1920 (9:16); square or vertical uploads are treated as Shorts on YouTube, 16:9 is not.
- Import into Captions for auto captions and chat-driven edits; Max ($24.99/mo, iOS price) adds 100+ caption templates.
- Cut a visual change every few seconds with B-roll, screenshots or text cards made in Canva.
- Remove watermarks by generating on a paid plan; HeyGen lists watermark removal on Creator and up, not Free.
Tool: Captions · Free option: canva (free video editor) or capcut
Tip: Keep the C2PA metadata from your AI tool intact on export; TikTok, YouTube and Meta read it to auto-label, and an automatic label applied from metadata cannot be removed on YouTube.
Upload with the right AI disclosure
A realistic avatar of a real person saying a script is the type of content all three platforms ask you to label, and missing it can cost you monetization.
- YouTube: in Studio upload, open the Attributes section, set 'AI use' (altered or synthetic content) to Yes for a photorealistic avatar.
- TikTok: turn on the AI-generated content label in the upload's more options before posting.
- Instagram/Facebook: switch on the AI label when posting; Meta shows 'AI info' on disclosed or detected content.
- Add a line in the caption such as 'Presented by an AI avatar of [name]' for viewers and clients.
Tool: HeyGen · Free option: Disclosure settings are free on every platform
Tip: YouTube says disclosing does not limit reach or monetization eligibility, while repeated non-disclosure can lead to removal or YPP suspension, so label every avatar upload by default.
What it costs
HeyGen Free gives 3 videos/month up to 1 minute each with watermark removal not included; Synthesia Basic gives 10 minutes and 9 stock avatars; ElevenLabs Free has about 10 minutes of TTS with no listed commercial license. Fine for testing, not for client work.
HeyGen · Synthesia · ElevenLabs · Canva
Low end is HeyGen Creator billed annually ($24/mo, no watermark, 1 digital twin). Adding ElevenLabs Starter ($6) and Captions Max ($24.99, iOS) brings it to about $55; HeyGen Pro instead of Creator raises it to about $80. Synthesia Starter works out to roughly $1.44-$2.40 per generated minute (12 min for $29 monthly; about 150 min a year for $18/mo annually).
HeyGen · ElevenLabs · Captions
| HeyGen | Creator | $29/mo monthly or $24/mo billed annually; 600 credits |
|---|---|---|
| HeyGen | Pro | $49/mo; 1,000 credits |
| Synthesia | Starter | $29/mo monthly or $18/mo billed annually |
| Synthesia | Pro | $89/mo monthly or $64/mo billed annually; about 60 min/mo, 5 personal avatars |
| ElevenLabs | Starter | $6/mo list price; commercial license, about 30 min TTS, Instant Voice Cloning |
| Captions | Max | $24.99/mo (iOS price) |
Rules to know before you post
- YouTube's altered or synthetic content disclosure ('AI use' in Studio Attributes) is required when realistic content makes a real person appear to say or do something they did not, which covers a photorealistic digital twin reading a new script.
- TikTok's Community Guidelines on AI-generated content require labeling realistic AIGC, and TikTok auto-labels uploads carrying C2PA Content Credentials.
- Meta requires an AI label on photorealistic video or realistic-sounding audio that was digitally generated or altered, shown as 'AI info', and may penalize accounts that skip it.
- HeyGen and Synthesia both require the depicted person to record their own consent for a custom avatar; Synthesia requires live recording with a passcode, age 18+, and bans avatars of celebrities and politicians.
Common mistakes
- Generating the full script before testing a short line: check pronunciation and lip-sync on 10 seconds first to avoid burning credits.
- Filming twin footage with cuts, music or a second voice: both HeyGen and Synthesia need one continuous take with only the subject's voice.
- Leaving the AI disclosure off because the avatar is 'just you': a realistic clone saying words you never recorded still counts as synthetic on YouTube and Meta.
- Exporting 16:9 and cropping later: generate the avatar in 9:16 so framing, gestures and captions fit 1080x1920 from the start.
Questions
Do I have to label AI avatar videos on YouTube Shorts?
Yes if the avatar is photorealistic and shows a real person saying something they did not say. Set 'AI use' to Yes in Studio's Attributes section; YouTube says this does not reduce reach or monetization eligibility.
How much does an AI avatar video cost per minute?
Synthesia lists about 12 minutes a month on Starter at $29 (around $2.40/min) or roughly 150 minutes a year at $18/mo billed annually (around $1.44/min). HeyGen bills in credits whose cost per minute varies by model, so run a test and check current pricing.
Can I make an avatar of my client or employee?
Only if they record their own consent in the tool. Synthesia requires a live consent recording with an on-screen passcode from the same person, and HeyGen requires the depicted person to record a consent video.
Is a stock avatar or a digital twin better for a coaching business?
A digital twin keeps your face and voice consistent with your brand, which matters when clients buy from you personally. Stock avatars suit product explainers and internal training where no single person is the brand.
Researched and checked on 2026-10-11. Sources: HeyGen Pricing · Create your first Digital Twin video avatar with Avatar IV (HeyGen Help) · HeyGen Avatar consent (developer docs) · Synthesia Pricing · How do I create my personal avatar from a video? (Synthesia Help) · How do I create a Studio Express-1 avatar? (Synthesia Help) · ElevenLabs Pricing · Captions Pricing · Disclosing use of altered or synthetic content (YouTube Help) · Upload YouTube Shorts (YouTube Help) · Partnering with our industry to advance AI transparency and literacy (TikTok Newsroom) · Labeling AI content (Meta Transparency Center) · Meta will start labeling AI-generated images on Instagram and Facebook (NPR). Prices and policies change, so confirm on the provider’s site.