Synthesia, ChatGPT, and the Hybrid Worship Service: A Practical Guide for Faith Leaders
Most congregations now run two services at once, the one in the room and the one on the screen, and faith leaders are being asked to make both feel present without doubling their staff. AI avatar tools like Synthesia and writing assistants like ChatGPT promise to help. The real question is not whether these tools work, but where they serve worship and where they quietly erode it. This guide walks through both.
The hybrid worship service is no longer an emergency measure. What began as a pandemic workaround has become a permanent fixture of congregational life, and for many faith communities the online audience now rivals or exceeds the number of people in the pews. That shift has created a steady, often invisible workload: someone has to run the stream, caption the sermon, cut the highlight clips, write the follow-up email, answer the questions that arrive in the chat, and translate the message for members who do not speak the dominant language of the service. In a small congregation, that someone is usually the same person who preached.
Into this gap step two very different kinds of AI tools. Synthesia generates video from a typed script, producing a lifelike avatar that speaks your words with synchronized lip movement, in dozens of languages, without a camera or studio. ChatGPT and similar assistants help with the writing-heavy work that surrounds a service: outlines, discussion questions, announcement copy, social posts, and translation review. Used well, they let a small team produce the kind of polished, accessible, multilingual experience that used to require a media department. Used carelessly, they introduce theological errors, replace human connection with synthetic presence, and erode the trust a congregation places in its leaders.
This article is written for faith leaders who are pragmatic rather than alarmist or starry-eyed. We will look at exactly what Synthesia and ChatGPT can do for a hybrid service, the workflows that hold up week after week, and the boundaries that protect worship from becoming a performance of automation. The recurring theme is simple: these tools belong in preparation and distribution, not in the sacred moments where embodied human presence is the entire point. For a deeper treatment of where that line sits, our article on AI and pastoral care is a useful companion.
What a Hybrid Service Actually Demands
Before reaching for any tool, it helps to name the work. A hybrid worship service is really two productions happening simultaneously, and the online one has needs the in-room one never had. The congregation in the building is carried by acoustics, eye contact, and shared physical space. The congregation on the screen is carried by audio clarity, captions, camera framing, and a sense that someone on the other side knows they are there. When churches treat the stream as an afterthought, the online community feels it immediately, and they leave.
The surrounding workload falls into a few predictable buckets. There is preparation, which includes study, outlining, and writing. There is production, which includes the live stream itself, captions, and translation. There is follow-up, which includes clips, devotionals, emails, and answering messages. And there is accessibility, which cuts across all three: language, hearing, vision, and reading level. AI tools map onto these buckets unevenly. They are strong at preparation and follow-up, useful but supervised in production, and genuinely valuable for accessibility. They are weak, and often inappropriate, in the live, relational core of the service itself.
The Four Workloads of Hybrid Worship
Where the time goes each week, and where AI realistically fits
- Preparation: study, sermon outlines, liturgy, discussion guides. Strong fit for a writing assistant used as a brainstorming and structuring partner.
- Production: live stream, captions, real-time translation. Useful with human oversight, never fully unattended.
- Follow-up: highlight clips, devotionals, recap emails, social posts. Strong fit for both avatar video and text generation.
- Accessibility: multilingual access, captions, plain-language summaries. High-value, mission-aligned use that genuinely widens the welcome.
Where Synthesia Fits, and Where It Does Not
Synthesia turns a written script into a video of a digital presenter. You can use one of its stock avatars or create a custom avatar trained on footage of a real person, and the platform will render that figure speaking your text in any of dozens of languages with convincing lip-sync. For a congregation, the appeal is obvious: announcements, welcome messages for newcomers, short teaching segments, and multilingual versions of the same content can be produced in minutes rather than requiring repeated filming sessions. A volunteer who would never agree to be on camera every week can write a script instead.
The honest, productive uses sit firmly in the periphery of worship rather than its center. A weekly announcements video that greets online viewers and points them to the right links. A short series of teaching clips that recap a sermon for people who missed it or want to share it. A welcome video translated into the languages your community actually speaks, so a first-time visitor hears about your ministries in their own tongue. Internal training for volunteers on how to run the stream or greet online guests. In each case the avatar is doing communication and logistics work, not standing in for a pastor in a moment of worship or care.
Defensible Uses
- Weekly announcement and welcome videos for online viewers
- Multilingual versions of logistical and informational content
- Volunteer and staff training segments
- Short, clearly-labeled teaching recaps and explainers
Lines to Hold
- No avatar preaching a sermon in place of a human in live worship
- No synthetic stand-in for prayer, pastoral care, or sacraments
- No avatar of a person used without their clear, ongoing consent
- No undisclosed AI content presented as a live human moment
The reason for the boundary is not technophobia. Research on the first fully AI-led worship services found that what congregants missed most was embodiment and resonance, the sense of a present human being who shares the room and bears witness alongside them. An avatar can deliver information flawlessly, but it cannot be present in the way worship assumes. When a synthetic figure replaces the preacher, the basic human need for co-presence between congregation and minister goes unmet, and people notice the absence even when they cannot name it. The closer a Synthesia video gets to the heart of the service, the more it takes away rather than adds. Our discussion of boundaries on spiritual chatbots traces the same principle into conversational tools.
Where ChatGPT Fits in the Preparation Cycle
If Synthesia is a tool for the face of your online presence, ChatGPT is a tool for the words behind it. Most faith leaders who use it well treat it as a study and drafting partner rather than an author. Given a passage, a theme, or a series title, it can produce an outline with cross-references, suggest illustrations, generate small-group discussion questions, and draft the announcement copy and social posts that surround a service. It is fast, tireless, and genuinely helpful for the administrative writing that consumes a disproportionate share of a minister's week.
The critical discipline is to keep the human in the position of theological judgment. A language model predicts plausible text; it does not know your tradition, your congregation, or what is true. It will confidently invent a quotation, misattribute a verse, or smooth a difficult doctrine into something more comfortable and less accurate. That is why the strongest workflows use AI early, for structure and stimulation, and rely entirely on the human leader for the content that carries doctrinal weight. The sermon's claims, its handling of scripture, and its pastoral application should come from the person who will stand before the people and answer for them.
A Healthy Division of Labor
Let the assistant do the scaffolding; keep the substance human
A useful rule of thumb is that AI handles form while the leader handles truth. The tool can shape and accelerate; it should never be the final word on what your community believes or how it cares for people.
- Good fit: outlines, discussion questions, plain-language summaries, recap emails, social captions, title brainstorming
- Supervised fit: translation drafts, reading-level adjustments, accessibility rewrites, all reviewed by a fluent or qualified human
- Keep human: doctrinal claims, scriptural interpretation, pastoral counsel, prophetic or prayerful content, anything spoken as your own conviction
One more caution belongs here. Whatever you type into a consumer AI tool may be retained and used to improve the model unless you have configured it otherwise or use an enterprise tier. Treat pastoral details, names, prayer requests, and any confidential congregational information as you would treat them in any other third-party system: keep them out. The convenience of pasting a real situation into a chatbot for advice is rarely worth the privacy cost, and it can violate the trust at the center of pastoral relationships. Our broader guide for faith-based nonprofits and AI covers these data-handling habits in more depth.
The Accessibility Case: Where Hybrid AI Earns Its Keep
If there is one area where these tools clearly advance the mission rather than just trim the workload, it is accessibility. A hybrid service is, at its best, a wider door. AI-driven captioning makes the sermon readable for members who are deaf or hard of hearing and for the many viewers who watch with the sound off. Real-time translation platforms now let a single spoken sermon reach attendees in a hundred languages, often through a phone app or a scanned code, with some systems even preserving a sense of the speaker's own voice. Plain-language summaries make dense theological content reachable for younger readers, newcomers, and people with cognitive disabilities.
These uses deserve enthusiasm precisely because they extend hospitality. A congregation that has always wished it could welcome its immigrant neighbors, or include a member who lost their hearing, can now do so at a cost that was unthinkable a few years ago. The pairing is natural: ChatGPT or a dedicated translation engine drafts the multilingual text, and Synthesia can render a welcome or explainer video in those same languages, so the welcome is not only translated but visible.
Even here, supervision matters. Machine translation of religious language is improving quickly but still stumbles on idiom, doctrine, and the weight of specific theological terms. A word that is ordinary in one language can carry unintended doctrinal freight in another. The safe pattern is to generate the draft with AI and have a fluent, ideally theologically literate, human review it before it represents your community. The goal is a real welcome, not a confident mistranslation. For congregations weighing where translation review needs a human in the loop, the same logic applies that we describe in our piece on honest disclosure when people interact with AI.
Accessibility Wins Worth Prioritizing
- Live and recorded captions for hearing accessibility and silent viewing
- Multilingual sermon access for members who do not speak the dominant language
- Plain-language summaries for newcomers, youth, and varied reading levels
- Translated welcome and explainer videos that make the invitation visible
A Practical Weekly Workflow
Theory becomes useful when it fits into the rhythm of a week. The workflow below shows how a small team can use both tools without letting either drift into the parts of worship that should stay human. The pattern keeps AI in preparation and follow-up, keeps the live service fully human, and builds in review at every point where theology or a person's likeness is involved.
Early Week: Preparation
The preacher studies the passage first, then uses ChatGPT to pressure-test an outline, surface cross-references, and draft discussion questions for small groups. Every theological claim is the leader's own. The assistant accelerates structure; it does not supply conviction.
Midweek: Logistics and Welcome
A volunteer writes scripts for the weekend's announcements and a newcomer welcome, then renders them with Synthesia, including translated versions for the community's main languages. These are clearly informational videos, never a substitute for a person leading worship.
Sunday: The Live Service Stays Human
Preaching, prayer, communion, and pastoral moments are led by people, in the room and on camera. AI runs quietly in the background as captions and, if used, real-time translation, with a human monitoring quality. No avatar appears in the worship itself.
After Sunday: Follow-Up and Reach
ChatGPT drafts the recap email, devotional series, and social captions from the sermon transcript; a human edits for voice and accuracy. Highlight clips and short teaching recaps extend the message through the week. This mirrors the repurposing approach in our guide to AI sermon repurposing for multi-site ministries.
Disclosure, Consent, and a Simple Policy
The fastest way to lose a congregation's trust is to let them discover, rather than be told, that a message they thought was personal was machine-generated. Research on AI in churches consistently finds that authenticity and transparency rank among the top concerns members raise, with many believing they have a right to know when AI helped produce their pastor's words. Yet only a small fraction of churches have any formal AI policy at all. That gap is where avoidable damage happens.
Disclosure does not have to be heavy-handed. A brief note that a video is AI-generated, a line in the bulletin describing how your community uses these tools, and a clear internal rule that AI is never presented as a live human moment will carry you a long way. Consent is the companion principle: if you create a custom avatar of a real person, that consent should be explicit, documented, and revocable, and it should never extend to having the avatar say things the person would not say. The likeness of a pastor or volunteer is theirs, not the institution's to repurpose indefinitely.
A workable policy fits on a single page. Name where AI is welcome, name where it is forbidden, require human review of anything theological or translated, require disclosure of AI-generated media, and require consent for any use of a real person's likeness or voice. Denominations are increasingly issuing guidance of their own, and aligning with it saves individual congregations from reinventing the wheel; our overview of how denominational bodies are writing AI policies is a good starting point.
Disclose
Label AI-generated video and clearly communicate how your community uses these tools, so trust is never a surprise.
Consent
Get explicit, documented, revocable permission before creating an avatar of any real person, and honor the limits they set.
Protect
Keep worship, sacraments, and pastoral care human. Use AI to widen the door, never to fill the seat that belongs to a person.
Conclusion
Synthesia and ChatGPT are not the future of worship, and they were never meant to be. They are production and communication tools, and the faith leaders who get the most from them are the ones who keep them in that lane. Let the avatar handle the announcement video and the translated welcome. Let the writing assistant draft the outline, the recap email, and the discussion questions. Then put both tools down when the service begins, because the part of worship that matters most is the part no model can supply: a present human being, sharing the room, bearing witness alongside the people they serve.
The hybrid service will keep growing, and the workload that comes with it is real. AI can absorb a meaningful share of that load, freeing staff and volunteers to spend their attention on people rather than production. The discipline is to spend the time you save on presence, not to use the tools to manufacture a presence that is not there. Used that way, with disclosure, consent, and a clear sense of where the line sits, these tools can make a small congregation feel larger, warmer, and more accessible than its budget should allow.
Start small. Pick one workload, perhaps the multilingual welcome or the weekly recap, and build a supervised workflow around it before adding more. Write the one-page policy before you write the first script. And keep returning to the question that should govern every choice: does this tool help us be more present to people, or does it quietly let us be less? Hold to that, and the technology serves the ministry instead of the other way around.
Bring AI Into Your Ministry Thoughtfully
We help faith-based organizations adopt AI in ways that extend hospitality and ease workload without compromising the human heart of worship. Let us help you build a workflow and a policy that fit your community.