A video platform for AI presenters, personal avatars, scripted scenes, voice direction, editable Agent drafts, and localization.

HeyGen is a video creation platform built around AI presenters, personal avatars, scripted scenes, and video localization. You can create a speaking video from a script, develop a Digital Twin from authorized footage, assemble visual material in AI Studio, or use Video Agent to develop a more complete video from a brief. Translation tools provide another route for adapting existing spoken content.
It is useful for recurring explanations: training modules, customer instructions, course introductions, product messages, and multilingual communications. The presenter can remain available while the script changes, which is a different workflow from arranging a new filming session for every update. The creator still needs to review the spoken message, delivery, scene design, captions, and any translated version before it reaches the audience.
AI Studio organizes the video around a script, avatar, scenes, voice, and visual elements. You can start with a blank project, a template, supported presentation material, a photo-to-video route, or Video Agent. Landscape and portrait orientations serve different destinations, while text and other scene elements can be adjusted in the editor.
Build a scene around one part of the message. For an onboarding lesson, the presenter can introduce the task, followed by a screenshot or diagram that explains the action. Keep the script aligned with what is visible. A speaker saying “select this option” needs the relevant option on screen, not an unrelated decorative background. Use a new scene when the explanation changes subject rather than relying on the presenter to carry every visual detail.
Preview the layout at the likely viewing size. A title that is readable on a laptop can become crowded in a vertical mobile version. Leave room for captions and avoid placing the avatar over an essential diagram label. Consistent scene structure also helps with future updates: a revised instruction can replace one scene instead of forcing changes throughout the entire lesson.
HeyGen provides supported stock avatars and custom-avatar workflows. A video-based Digital Twin uses footage of the person represented. Its creation process requires that person's consent recording; if the avatar is for someone else, that person supplies their own consent. The verification step belongs to the product workflow, alongside preparing the source footage.
Choose the presenter around the audience and purpose. A familiar company representative may suit recurring internal messages; a stock presenter may be sufficient for a neutral explanation. The video should be clear about who is communicating and should use a script that person or organization is comfortable presenting. The convenience of a reusable avatar makes script review more important, because a new message can be created without a new performance recording.
The filming guide provides guidance on clear footage, steady framing, lighting, and controlled movement. Review a brief test recording before completing the source take. A distracted expression, unstable focus, or unclear sound can become part of the material used to create the avatar. Prepare the input for the actual style of presentation you intend to produce.
Avatar V uses video footage to guide a person's movement and delivery. HeyGen's current input guide distinguishes video-based human looks from photo-based looks and virtual or non-human characters. Supported engines and motion settings differ, so choosing an avatar is also a choice about the appropriate source material.
Looks provide variations in appearance, outfit, background, or framing while retaining the underlying avatar identity. A Look and a Voice can be changed separately in supported workflows. That helps adapt the presentation to different contexts without treating every visual variation as a new person.
Keep a small set of approved Looks for recurring communication. For example, use one for formal training and another for a brief social introduction. Check the first output from each combination of Look, engine, and voice before building a long series around it. A suitable face alone does not establish that the framing, gestures, and delivery fit the message. Review the combination as the audience will see it.
HeyGen supports script-based voices and eligible voice-cloning workflows. Voice Mirroring and Voice Director provide ways to shape delivery beyond ordinary typed punctuation. The documented mirroring process uses recorded or uploaded speech to communicate pacing and expression, while voice-direction tools provide another way to adjust how a line is presented.
Use delivery to support the meaning of the sentence. A welcome message can sound conversational, while a technical instruction needs clear pacing and separation between steps. Avoid making every phrase equally emphatic. Record a short example if the desired rhythm is difficult to describe, and listen to the result before applying the approach to every scene.
Check names, abbreviations, numbers, and specialist terminology. A convincing voice can still pronounce an important product name incorrectly. Keep the script concise enough to follow alongside the visual material, and allow time after an instruction for the viewer to understand the result. Voice control helps communicate the text; it does not replace checking whether the text itself is correct.
Video Agent develops video from a brief, combining supported presenter material, visuals, and motion elements. The Agent FAQ describes revising through the conversation or editing a copy in AI Studio for manual control over text, timing, scenes, and layouts. Generation and revisions have mode-dependent credit implications.
Give Agent the structure of the message, not merely its topic. For a customer announcement, identify what changed, who needs to act, and the intended next step. Supply accurate product details and any visual material that must be used. Review the proposed or generated sequence for omissions before spending time polishing its appearance.
If the result has the right structure but one scene needs adjustment, identify that scene and the desired change. Move to AI Studio when exact positioning or timing matters. Keep the generated draft and the edited copy understandable as separate versions. That makes it easier to compare whether a visual improvement also altered the intended spoken message or removed an important instruction.
HeyGen supports PDF and presentation-based video workflows, with imported material providing scenes around which narration and avatars can be developed. This is relevant for training decks and educational material that already have a visual structure. The slides are a starting point for a video, rather than a guarantee that the original deck is ready for spoken presentation.
Review each imported scene as a frame. A slide designed for a live speaker may contain too much text for a video viewed on a phone. Shorten the visible wording and let narration explain the details. If an avatar is added, make room for it instead of covering the original slide. Check charts and screenshots at the size viewers will actually see.
Write a script that connects the scenes. Reading every bullet verbatim can be repetitive, while an unstructured voiceover can leave the viewer unsure which point is being discussed. Introduce the purpose of a chart, explain the relevant observation, and then move on. For a frequently updated course, keep each scene's script tied to the corresponding source material so a changed figure can be revised reliably.
The Video Translation guide distinguishes audio-only translation from engines that also generate lip synchronization. Input conditions, engine selection, language support, and proofreading access affect the workflow. Review & Edit allows eligible users to check the translated script before final generation. Translation of spoken audio does not automatically translate every visible label in the source video.
Begin with a clear source and a corrected understanding of what it says. A technical lesson may contain terms that have a preferred local equivalent, while a product name may need to stay unchanged. Have a fluent reviewer check the translated message and the way instructions are expressed. A phrase that sounds natural can still use the wrong term for the audience's actual interface.
Pay attention to timing. Translation may require more or fewer words than the original, and duration settings can affect the final video. Review the pace of each important instruction and the relation between speech and screen action. If the source shows a button for only a brief moment, a longer translated phrase may need a different visual edit to remain understandable.
The Brand Glossary supports pronunciation and translation rules. Those are distinct tasks: a pronunciation rule affects how a term sounds, while a translation rule can specify a preferred rendering or preserve a term. HeyGen recommends organizing language-specific rules where translations differ.
A practical glossary for a product lesson might include the product name, several feature names, and an abbreviation used throughout the course. Apply the rules and listen to the generated speech. Preview a term in the sentence where it matters rather than judging an isolated pronunciation alone. Keep the glossary small enough that reviewers understand which terms have special treatment.
Caption support differs by workflow. AI Studio offers editable styled captions, while translation sharing has its own closed-caption options. The current guide distinguishes those from toggleable closed captions for avatar videos. Check the actual output and sharing route needed by the audience. Correct caption wording and make sure text does not cover a diagram or the speaker's face.
A Brand System organizes visual assets such as logos, colors, fonts, motion elements, and chosen avatars. It can be applied in supported AI Studio and Video Agent workflows. Branding is separate from the glossary's language rules and from the script's factual content.
For a recurring video series, establish a restrained title style and approved presenter before creating many episodes. Keep information hierarchy consistent: task name, explanation, then the action viewers should take. A recognizable style is useful when it makes the content easier to follow. Large logos and motion graphics can become distractions if they compete with the instruction.
Team collaboration depends on access and workflow. Share a draft with reviewers at the stage where they can still change the script and the visual arrangement. Ask for specific checks: accuracy from the subject expert, terminology from the local-language reviewer, and framing from the producer. A single review of a polished final render is less convenient than resolving those questions in the editable project.
Suppose a team needs to explain how new customers activate a notification. Prepare an approved three-part script: where to find the setting, how to change it, and what confirmation to expect. Choose an appropriate presenter and build the scenes in AI Studio around actual screenshots. Keep the key controls visible and use narration to connect the steps.
Preview the delivery, correct the pronunciation of the feature name, and apply a glossary rule if it recurs. Add readable captions and inspect them against the screenshots. Have the product owner review the source version before localization. Translate the spoken lesson with the appropriate engine and proofreading route, then ask a fluent reviewer to check the target script and timing.
Replace visible interface text or screenshots where the localized audience uses a different version. Save the approved source and translated projects with the glossary and brand assets used. When the product changes, update the affected scene and repeat the relevant review instead of assuming the earlier translation remains correct. This process gives each HeyGen feature a specific role in a recurring communication task.
The pricing page includes free and paid individual access plus business and enterprise options. Plans differ in credits, video length, output resolution, custom avatars, translation review, collaboration, and other features. Check the selected avatar engine and generation mode as well as the plan name. A feature advertised for the platform may have its own allowance or eligibility conditions.
API access is separate from regular creative subscriptions. For normal production, budget for revisions and localization as well as the first generated video. Avatar motion, pronunciation, timing, translation, and captions still need review. A personal avatar can present a prepared message; it does not make the script an authentic recording of something the person said in a live event.
The person represented must provide the required consent recording, alongside suitable footage. Plan that step with them before starting the avatar workflow.
Spoken translation and visible scene content are separate. Review screenshots, charts, titles, and interface labels for the localized audience.
The documented workflow includes editing a copy in AI Studio. Use it for precise adjustments and review the resulting script and layout together.
No. The current guide distinguishes supported video-based human looks from photo-based or virtual-character workflows. Select the engine that matches the input.