Weekly spokesperson videos for a client: how an agency plans it

A client has asked for a spokesperson video every week. Booking a real presenter that often is costly, and the retainer was priced on the assumption that it would not be. This page is for the agency owner who has to decide how to deliver it, what it really costs, and what to put in the contract.

I use Higgsfield, an online AI image and video platform, and I will say plainly where it fits and where it does not. For a plain weekly presenter it is often not the best tool.

This page may contain affiliate links from which Blur Studio earns commissions.

The problem: a client wants a face every week

"Spokesperson" can mean three different things. Ask the client which one they picture, because the cost and the rules differ.

• A stock avatar. A licensed AI presenter from a platform's library. Fast and cheap, but other brands can use the same face.
• A custom avatar of a real person. Usually the client's own founder or a staff member, recorded once and then driven by scripts. This needs written consent.
• A synthetic character. A made-up person or mascot that exists only as AI output. This is where the advertising disclosure rules are strictest.

How it works in plain words: you pick or build the presenter, paste a script, choose a voice and a language, and the tool returns a video of that presenter speaking it. Most avatar platforms also offer a cloned voice (a synthetic copy of a real person's voice, which needs the same consent as the face) and translation, so one script can become several language versions.

Four ways to do it, cheapest first

1. Voice note plus slides. Someone on the client's team records a two-minute voice note each week over a few slides. No face, no consent question about likeness, and no AI cost. If the client's real goal is "a regular update", this may be enough.
2. A monthly batch shoot. Book a real presenter for one half-day and film four to six videos from scripts written in advance. You pay talent once a month, not weekly. For authentic emotion, a reaction to events, or any health, money or legal topic, this beats an AI presenter.
3. An avatar platform. HeyGen, Synthesia or Creatify, covered below.
4. A freelance creator. Ask for a quote for four short videos a month. Prices vary too much for me to quote a number.

One study cited in a guide from Visla (a video software company, July 2026) found that viewers rated human presenters somewhat better, while exam results after the videos were similar. Take that as a reminder that an AI presenter can cost you some warmth, even when it teaches as well.

What a weekly cadence really costs

The software fee is the small part. Count four things for each video: writing the script, the credits (a credit is the unit a platform uses to meter how much you can make), retakes, and review time. Review time is the one agencies forget to price.

Here is a worked example with made-up numbers. A client wants one 60 second video a week, four a month.

• Scripts: 1 hour to write four, 30 minutes for the client to approve them.
• Generation: say 100 credits a finished minute on Synthesia, as its pricing FAQ describes (more if the avatar is on screen throughout), so about 400 credits for the month. Add 50 percent for retakes and you are near 600.
• Review: 15 minutes per video for you, 10 for the client.
• Delivery: add the AI label (see below) and captions, about 15 minutes each.

Total people time is about 4 to 5 hours a month. Compare that with one monthly batch shoot day. If your edit and review time per AI video climbs above what a shoot day costs spread over four videos, the AI route has failed on cost. The pilot below tests exactly that.

Avatar platforms compared for weekly use

These are vendor prices as read on 3 November 2026 unless I say otherwise. They change often, so check each pricing page before you quote a client.

Platform	What it is	Plans I could confirm	Notes for weekly use
HeyGen	AI avatar and video platform	Free $0 (3 watermarked videos a month, up to 1 minute); Creator $29 a month (600 credits); Pro $49 a month (1,000 credits)	Creator includes a custom avatar with your own voice and likeness and 175+ languages. Credits roll over one month on Creator. Business is listed at $149 a month (1,500 credits, videos up to 60 minutes, shared workspace, SCORM export). HeyGen's page says credit use varies by model, so test before you promise a volume.
Synthesia	AI video platform with avatars, used a lot for training and business video	Basic $0 (500 credits a month, watermark); Starter $29 a month (first month $14, 1,250 credits a month) or $18 a month billed $216 yearly (15,000 credits a year); Pro $89 a month or about $59 to $64 a month billed yearly (72,000 credits a year)	Video length up to 3 minutes on all plans. A typical finished minute uses about 100 credits (more if the avatar is on screen throughout), and you are charged only for new avatar footage. Synthesia showed different prices on different page loads, so check the live page.
Creatify	AI tool for making video ads with AI actors	Free plan: 10 monthly credits, 300 AI actors, watermarked exports. Paid prices read on 3 November 2026: Standard $15 a month ($10 on annual billing, 55 credits a month), Pro $55 ($38 annual, 280 credits)	Built around ads more than talking-head series. Credits expire after 2 months.

For a head-to-head on one of these, see my Higgsfield vs HeyGen page, and for Creatify Higgsfield vs Arcads and Creatify.

Pick an avatar platform if you need a presenter for 30 to 180 seconds each week, want templates, or need a team workspace. Pick it for a custom avatar of a real person with a consent flow built in.

Where Higgsfield fits, and where it does not

Higgsfield's Lipsync Studio makes a still image or a clip talk from typed text or uploaded audio. The models (the AI engines behind the tool) include Higgsfield Speak 2.0, Infinite Talk, Kling Avatars 2.0, Wan 2.5 Speak, Kling 2.6 Lipsync and Veo 3.1. I cover the feature in my Higgsfield Lipsync Studio guide.

The catch for a weekly presenter is length. The pricing page prices these models per short clip. For example, Speak 2.0 at 720p is listed at roughly 14 credits for a 5 second clip (a vendor figure, approximate). Higgsfield's help center describes Infinite Talk as suited to long-form talking video and Kling Avatars 2.0 as giving longer clips, but I could not find a length limit in the text I read, so check the limit on screen. Higgsfield's own guide advises keeping each generation under 30 seconds and cutting between clips. A 60 second presenter video made from stitched clips means you are the editor, matching lighting, eye line and expression at every cut.

Example with made-up numbers: a 60 second video is twelve 5 second clips. At about 14 credits each that is about 168 credits per video, and about 670 for four. With 50 percent for retakes you are near 1,000 credits a month. Plus is listed at $59 a month ($47 billed annually) with 1,200 credits as of November 2026, so one client could use most of a plan. Check the pricing page, and see Higgsfield pricing explained. Higgsfield's compare table lists Speak as included on Plus and Ultra only.

Pick Higgsfield if the presenter is stylised or not human (a mascot, an animal, a character), you want a 10 second hook or product intro, or you already pay for Higgsfield for other work. If the client wants a plain human-looking presenter every week, HeyGen or Synthesia are purpose-built for that, and I would use them. If the brand has no face to show, see seven ways to add a human feel. There is no independent evidence I found that Higgsfield beats them at that job. To give a training video a brand mascot, see reusing a training video with a brand character.

Set up the presenter

• Stock avatar: pick one and write down which. Check whether the client's competitors use the same face.
• Custom avatar of a real person: get a signed agreement before recording. HeyGen's moderation policy requires the explicit consent of the person being represented and bars avatars of other real people, public figures included, without it. Write down the scope (which videos), the term, and what happens if the person leaves the client. Set the end date before you record.
• Character: keep a one-page character sheet with reference images and the words the character never says.
• Pronunciation lexicon: a file of client names, places and product terms, with phonetic spellings pasted into each script. Most embarrassing AI presenter errors are brand names said wrong.

The weekly workflow: four scripts, one review, one delivery

1. Agree the presenter and the script format. Same structure every week: hook, one point, one next step.
2. Write four scripts at once in one 90 minute session. This avoids the Friday rush that makes teams skip review.
3. One reviewer reads all four against the client's approved claims sheet.
4. Make the videos together.
5. Check for errors: names, numbers, lip sync, eyes, hands, and captions.
6. Add the AI label to the template (see next section) so it is never forgotten at 5 pm on delivery day.

Consent, disclosure and contract terms

This is general information, not legal advice. Have a lawyer look at your contract, and check local law and each platform's current rules.

• Disclosure in New York. A law firm summary (Cooley, January 2026) says New York's synthetic performer ad law, signed in December 2025, was set to take effect on 9 June 2026, so check that it is in force. It asks for a clear notice in the ad itself, applies where the advertiser has actual knowledge, carries civil penalties of $1,000 for a first violation and $5,000 for each later one, and exempts audio-only ads and AI used only to translate a human performer. It is aimed at digital performers that do not look like an identifiable real person, so a made-up presenter is the core case and a recognisable employee's avatar is a different question. Cooley advises tightening agency and vendor contracts to say who is responsible for the notice.
• Disclosure in California. Law firm alerts (Manatt and BakerHostetler) report that California's SB 1050, signed on 16 September 2026, takes effect on 1 January 2027. It covers audio, video and audiovisual ads that prominently feature a synthetic performer, audio-only ads included, suggests wording such as "this performance features a synthetic performer", and may allow private lawsuits. Read the statute before you advertise there.
• Platform labels. TikTok's guidelines (effective 24 September 2026) ask creators to label AI-generated or significantly edited content that shows realistic people. YouTube requires disclosure of realistic altered or synthetic content. Meta requires its AI disclosure tool for organic posts with photorealistic video or realistic audio. Check each platform on the day you post.
• FTC. The US Federal Trade Commission (FTC) is the federal consumer-protection regulator. Its influencer guide covers paid or gifted endorsements, not AI as such: the disclosure goes with the message, inside the video itself and not only in the description or a pile of hashtags, and nobody may claim experience with a product they have not tried. Law firm summaries say the same Endorsement Guides apply to virtual endorsers. Treat the AI label and the sponsorship label as two separate things.
• Higgsfield's Terms. The Terms of Use list representing an Output as human-generated when it is not as a prohibited use. Read the current wording before you deliver client work.
• No testimonials. Never have the spokesperson say "I use this every day". Keep the character to explaining, announcing and showing.
• Contract clauses to consider. Who supplies the avatar. The written scope of use for any real person's face and voice. Term and takedown when they leave. Who adds the disclosure. An approval turnaround time. The agency's right to label content as AI. A line that outputs are not guaranteed to be unique.
• Governance. Visla's published checklist is a usable start: name an owner for each avatar, say who approves scripts and finished videos, and list what the avatar may never say (apologies, layoffs, safety incidents and personal testimony should stay with the real person).

The free-account rule matters too. Higgsfield's help center says every free-account output carries a watermark. Whether free outputs may be used commercially is unclear: the pricing table lists commercial use as not included on Free, while the help center says use of outputs is not restricted to paid plans. I would use a paid plan for client work, and see Higgsfield free vs paid.

A four-week pilot with pass and fail marks

This is a plan you can run before you sign a retainer for the whole year.

• Week 1: make one video on two platforms from the same script.
• Weeks 2 to 4: make one video a week on the better platform, with the full workflow above.
• Score: the client rates each video 1 to 5 for likeability, trust and "would you publish". You also log the real credits and hours spent.

Pass means all of these: the average score is 3.5 or better, there are zero name pronunciation errors after one fix, the credits stay within the budget you set, and your hours per video are lower than a monthly shoot day would be per video. Fail on any one and offer the batch shoot instead. Use the real credit figures from the first five videos in the retainer, not guesses.

Who should skip AI spokespersons

• Clients in health, money or legal topics where trust is the product.
• Any message that needs real emotion: apologies, news, crisis updates.
• Clients whose audience already distrusts AI content.
• Agencies without time for the weekly review. Skipping the review is where the damage happens.

Pick this if

Situation	Pick
Update only, no face needed	Voice note plus slides
Real warmth matters	Monthly batch shoot
Plain presenter, 30 to 180 seconds weekly	HeyGen or Synthesia
Ad-style videos with AI actors	Creatify
Stylised character or short hook	Higgsfield Lipsync Studio

How I'd approach this

I would price the retainer on the batch shoot, offer the AI presenter as a cheaper tier, and run the four-week pilot before promising weekly delivery. Put the AI label in the template, keep one lexicon file per client, and record real credits per video so the next quote is based on facts. For a deeper look at keeping one character steady, see my page on a recurring AI character for social. If you run this for several clients, Blur Studio builds the repeatable workflow, approval portal or shared asset library that holds the scripts, the sign-offs and the labels in one place.