Talk to Me Doja AI Trend: The Cheek-Sway Reaction

The performer plays it cool, presses both palms to the cheeks, then sways into a shy grin.

  • Photo to video
  • 13 s video
  • 1 photo

See the effect

Effect preview

Effect preview: 13.8 s. This preset creates a 13 s vertical video.

The Talk to Me Doja AI Trend is a reaction-dance format pairing a composed opening with palms-to-cheeks contact, a head tilt and a bashful side-to-side sway. Also called Talk to Me Doja AI dance · Speak to me Doja · Doja talk to me · Doja Cat cheek sway · Urara Shiraishi dance · Urara Shiraishi AI trend.

ViralVideoAI editorial guide · Updated

What the Talk to Me Doja AI Trend video looks like

The effect preview illustrates the visual format. The scene plan below describes this page’s personalized generation preset. This preset creates a live-action-style adult performance. It does not turn the person into an anime character, add the anime GIF or supply the Doja Cat recording. The animation reference belongs to the format’s history, while your output uses your uploaded identity.

Watch the playable effect preview above, then compare it with the generation sequence below. Effect preview: 13.8 s. This preset creates a 13 s vertical video.

How to make Talk to Me Doja AI Trend on ViralVideoAI

Create the scene from one photo; the setting and motion are supplied by the preset.

  1. 1. Choose your photo

    Use one adult waist-up or thigh-up photo, with the face, hands and shoulders visible and unobstructed.

  2. 2. Generate the scene

    Sign in and choose Generate. The preset stages a 13-second vertical video; follow progress here or in My videos.

  3. 3. Review and download

    Check identity, the central action and the ending. Download the completed file, then add captions or an available licensed soundtrack if desired.

Which photo works best

Use one adult waist-up or thigh-up photo, with the face, hands and shoulders visible and unobstructed.

  • JPG, PNG or WebP · up to 10 MB each. One clear subject per image.
Upload photos and generate ↑

Where did the Talk to Me Doja AI Trend start?

The following dated notes distinguish the older format, individual posts and later AI adaptations. Figures are attributed reports, not live platform measurements.

Dated social audience figures below are attributed reports from MakeViral’s source notes. Original post links are supplied for checking; primary social counters were not independently verified. Wider tags, sound uses and individual posts measure different populations. Research review: . Read MakeViral’s source notes.

Post or snapshot dateAttributed historical note
2026-11-07Know Your Meme’s entry identifies @lovingbalanci’s Speak to Me Doja post as its earliest known example, not a proven first invention. MakeViral separately reports 247,400 views by 2026-11-13.
2026-11-09Know Your Meme records a human recreation by @the_real.ryan at over 3.1M views within three days. MakeViral’s later snapshot reports 3.3M views and 420,500 likes by 2026-11-13; the cutoffs differ.
2026-11-10@twin.cities.mattress’s warehouse recreation was reported by MakeViral at 605,900 views and 86,100 likes by 2026-11-13.
2026-11-13Reported tag: #talktomedoja 631 videos / 11.96M views. Know Your Meme added its entry on 2026-11-12; the primary social counters were not directly verified here.

Original posts and further reading

What happens, second by second

This is the intended sequence for the generation preset. It is a scene plan, not a frame-accurate transcript of the example. Timing and small gestures may vary.

TimeWhat happensWhy this beat matters
0–3 sLean toward the lens with a composed expression.The restrained opening establishes the contrast.
3–6 sStep back slightly, glance sideways and settle the shoulders.Leave room beside the face for both hands.
6–8 sRaise both palms together to rest gently against the cheeks.Keep all fingers separate from facial features.
8–11 sTilt the head and sway the upper body with a shy smile.The cheek-sway movement is the recognizable payoff.
11–13 sComplete the sway and release the hands.The face remains recognizable after the brief occlusion.

What to input to recreate it

Use one adult waist-up or thigh-up photo, with the face, hands and shoulders visible and unobstructed. This preset creates a live-action-style adult performance. It does not turn the person into an anime character, add the anime GIF or supply the Doja Cat recording. The animation reference belongs to the format’s history, while your output uses your uploaded identity.

Input decisionWhat works for this scene
Number and role orderone photo: Your adult photo. Keep one subject in each image so identity assignment is clear.
Face and framingUse one adult waist-up or thigh-up photo, with the face, hands and shoulders visible and unobstructed.
Clothes and settingThe preset builds this setting: A bright ordinary room with a single adult performer framed from the thighs upward, using a fixed phone-camera view. An exact matching location is unnecessary.
File qualityJPG, PNG or WebP up to 10 MB each. Use the original file with natural detail rather than a filtered screenshot.
User controlsUpload the labeled photos and generate. This preset has no public prompt field or additional scene selector.
Output and sound13-second vertical 9:16 video at 720p. New scene audio; add a specific licensed track in your publishing app if needed.

What goes wrong, and how to fix it

ProblemLikely causeWhat to try
Hands fuse with the cheeksThe source hides fingers or strongly compresses the face.Use a clear photo with separated hands and check the contact beat.
The head stretches during the swayThe input uses a distorted wide-angle selfie.Use a natural portrait taken farther from the face.
The result becomes an anime characterThe reference is an illustration or GIF screenshot.Upload the real adult portrait intended for the live-action performance.
The face or markings drift after a movementThe reference is blurred, strongly filtered or mostly in profile.Use a sharp, evenly lit image and compare identity around 6–8 s, when this scene’s central action occurs.
The opening and ending do not connect clearlyA movement or cut differs from the intended sequence.Review the transition from “Lean toward the lens with a composed expression” to “Complete the sway and release the hands”. Choose a simpler source pose and avoid adding a collage.
A caption or small printed detail is garbledGenerated lettering is not a reliable typesetting method.Keep the source free of large text overlays. Add captions and any required labels after downloading, using a normal editor.
The middle beat (6–8 s) feels rushed or unreadableThe action “Raise both palms together to rest gently against the cheeks” depends on a clear starting pose that the source photo does not establish.Re-shoot or re-crop so the subject reads at phone-screen size before 6–8 s, then regenerate. Do not speed up in an editor — the beat timing is already set by the preset.
The result crops the subject at the knees or shouldersThis preset frames a single-subject scene around the labeled upload; a tight headshot gives the model no body to animate.Use one adult waist-up or thigh-up photo, with the face, hands and shoulders visible and unobstructed. Re-upload with the required framing — do not enlarge a small image to fake extra body room.

Which caption and hashtags fit

Caption idea: “Trying to act normal for six seconds.” Use Talk to Me Doja AI dance as an alternate name when it fits the audience.

  • Suggested tags: #talktomedoja #speaktomedoja #cheeksway.
  • Choose relevant descriptive tags rather than unrelated popular ones.
  • Review identity and audio before sharing; add music through an available licensed library.

Pricing / What’s included

The preset creates one personalized video with the scene, length and photo roles described above. Review the current plans and packs before generating. Follow task progress in your account and download the completed result.

Talk to Me Doja AI Trend questions

What is the Talk to Me Doja AI Trend, and what makes its visual format recognizable?

The Talk to Me Doja AI Trend is a reaction-dance format pairing a composed opening with palms-to-cheeks contact, a head tilt and a bashful side-to-side sway. The scene is described as follows: A bright ordinary room with a single adult performer framed from the thighs upward, using a fixed phone-camera view. The sequence starts with this instruction: Lean toward the lens with a composed expression. Its payoff is equally specific: Complete the sway and release the hands. Those visual relationships distinguish the format from an unrelated clip that happens to use the same sound. Alternate search names include Talk to Me Doja AI dance, Speak to me Doja, Doja talk to me, Doja Cat cheek sway, Urara Shiraishi dance, Urara Shiraishi AI trend. This page describes both the wider format and the particular photo-to-video adaptation available in the upload panel.

How can I create the Talk to Me Doja AI Trend from photos without recording the performance?

Use one adult waist-up or thigh-up photo, with the face, hands and shoulders visible and unobstructed. Add one photo to the labeled upload area, sign in and choose Generate. The preset supplies the setting and staged motion, so a filmed performance is unnecessary. Wait for the task to finish, then review the beginning, middle and ending before downloading. The intended output is a 13-second vertical video. You do not need to find a CapCut project, paste a social-post URL or write the production instructions yourself. A successful upload is the starting reference, not a finished video.

Which starting photo gives the clearest identity in this particular scene?

Use one adult waist-up or thigh-up photo, with the face, hands and shoulders visible and unobstructed. Prefer the original image to a compressed social screenshot. Look for visible eyes, a distinct outline and natural facial detail; avoid strong beauty filters that erase the features the animation needs to preserve. In this scene, the important identity check occurs around 6–8 s: Raise both palms together to rest gently against the cheeks. Compare that moment with the opening rather than judging only a distant wide shot. If you have several possible photos, choose the one whose face is easiest to recognize at phone-screen size.

How many photos should I upload, and does their order affect the roles?

This preset uses one photo. The role labels are Your adult photo. A single clear subject is the useful input; adding a collage or a group screenshot makes identity assignment ambiguous rather than providing extra scene control. The setting and supporting action are supplied by the preset. Extra faces in the source can become competing references.

What happens in the opening seconds, before the main reveal or reaction?

The intended opening covers 0–3 s. Lean toward the lens with a composed expression. The restrained opening establishes the contrast. This setup matters because a viewer needs to understand the initial arrangement before the next change can mean anything. When reviewing your result, check whether the lead is recognizable, whether the camera establishes the setting and whether the first action starts cleanly. Do not remove the entire setup simply to reach the reveal sooner: an unexplained close-up can weaken the visual joke or comparison even when the final frame looks attractive.

Which moment in the middle of the video carries the main visual change?

Around 6–8 s, the scene follows this beat: Raise both palms together to rest gently against the cheeks. Keep all fingers separate from facial features. The preceding shot creates the expectation that this moment changes. Watch the transition at normal speed first, then pause if an edge, face or prop looks unstable. A technically sharp frame is not sufficient if the action before and after it does not connect. For this format, the viewer should be able to explain the change in a short sentence without needing a long caption to repair the sequence.

How should the final shot look, and which frame makes a useful cover image?

The planned ending is 11–13 s: Complete the sway and release the hands. The face remains recognizable after the brief occlusion. A useful cover shows the face clearly and includes enough of the setting to identify the format. It need not be the most extreme gesture in the clip. Avoid choosing a frame halfway through an occlusion, a fast turn or a clothing change, when the face and hands are more likely to be partially hidden. Review the final held pose as one option, then compare it with the middle reveal at actual display size.

Does this preset reproduce every option or output format shown in other versions?

This preset creates a live-action-style adult performance. It does not turn the person into an anime character, add the anime GIF or supply the Doja Cat recording. The animation reference belongs to the format’s history, while your output uses your uploaded identity. The upload panel is the practical specification for this version: one photo, the labeled role assignment and a 13-second vertical result. A cultural format can have still-image, hand-edited and generated-video versions at the same time. An option mentioned in an older tutorial should not be assumed to exist here unless it appears as a control on this page. Read the scene table to see what the preset directs before you start.

Do I need the same clothes, props or location in my reference photo?

A matching filming location is unnecessary because the scene is generated: A bright ordinary room with a single adult performer framed from the thighs upward, using a fixed phone-camera view. The reference photo mainly guides the featured identity and visible appearance. Clothing and props can change where the preset explicitly stages them, so this is not a promise that every seam, printed logo or accessory will remain identical. Choose an uncomplicated source with the required face and body framing. If an exact product, garment construction or documented location must be reproduced, inspect the output closely rather than treating the scene as a factual product photograph.

Which file types are accepted, and should I crop or enlarge a small image first?

Use JPG, PNG or WebP, with each image no larger than 10 MB. A clean original is more useful than a small image enlarged until it looks smooth. Crop away unrelated people and excessive empty background while keeping the face and required body parts intact. Use one adult waist-up or thigh-up photo, with the face, hands and shoulders visible and unobstructed. Do not crop so tightly that hair, hands or feet needed for the movement disappear. Uploading a file within the size limit does not restore detail that was already lost to blur, compression or a heavily filtered screenshot.

Can I change the scene, choose a different ending or supply my own detailed prompt?

This page runs the described preset rather than a general scene editor. The planned opening is “Lean toward the lens with a composed expression”; the ending is “Complete the sway and release the hands”. The public scene breakdown explains the intended order without exposing the private production instructions. There is no additional text box here for replacing the story, selecting an unrelated prop or forcing an exact frame. If you want another type of movement, compare the related templates below and choose the scene whose built-in action fits your idea.

How closely will the generated movement match the second-by-second table?

The table is an intended sequence for the 13-second preset, not a frame-accurate editing timeline. Small gestures, camera speed and the exact point of a cut can vary between renders. The useful test is whether the ordered actions remain understandable: setup, the central change and the final reaction. In particular, check 6–8 s, where the intended action is raise both palms together to rest gently against the cheeks. If you need an exact cut on a musical beat, download the result and make that final timing adjustment in an editor.

What can I do when the face, hands or important prop looks wrong?

One scene-specific issue is “Hands fuse with the cheeks”. A likely contributor is: The source hides fingers or strongly compresses the face. Use a clear photo with separated hands and check the contact beat. Also compare identity at the start and finish, count visible hands where they cross a prop, and look for unexpected duplicate bodies. Correct the source problem you can actually see before trying another render. An unclear photo cannot be fixed by expecting a more elaborate scene. Artistic imperfections in a completed video are different from a task that fails technically, so review the result before deciding what to change.

Why can a clean-looking portrait still produce a confusing transition?

The issue “The head stretches during the sway” can happen when the input uses a distorted wide-angle selfie. Use a natural portrait taken farther from the face. A portrait may look attractive as a still while supplying little information about a turning face, separated fingers or a full standing body. Match the photo to the motion rather than choosing solely by flattering lighting. The central action is “Raise both palms together to rest gently against the cheeks”. The source should give that movement a clear identity anchor. Compare the relevant table row with the frame where the result becomes confusing.

Does the video include the original viral song or a recognizable performer’s voice?

The generation requests new scene audio rather than a licensed copy of a commercial recording or a clone of an identifiable performer’s voice. The visual sequence can therefore be understood separately from a named song. If you want a particular track, add an available licensed sound in the app where you will publish, subject to the options for your account and use. Listen to the generated file before sharing; a video preview having music does not establish that a specific commercial recording is included in your output.

Who started this format, and can the earliest linked post be called the original?

Know Your Meme’s entry identifies @lovingbalanci’s Speak to Me Doja post as its earliest known example, not a proven first invention. MakeViral separately reports 247,400 views by 2026-11-13. An earliest example found in a source’s research is not necessarily the first example ever posted. Reposts, deleted originals, sound uploads and later AI adaptations can have different dates and creators. The timeline preserves those distinctions and links the supporting posts when a direct URL was available. This page does not claim to have interviewed the creator or watched a private original. For attribution in your own writing, name the specific post, platform and date rather than assigning the entire format to an unverified first creator.

How popular is the trend, and what date do the reported audience numbers describe?

@twin.cities.mattress’s warehouse recreation was reported by MakeViral at 605,900 views and 86,100 likes by 2026-11-13. These figures are attributed historical snapshots from MakeViral’s notes, reviewed here on 2026-11-13; we have not independently verified the primary counters. A post’s views, a sound’s uses and a hashtag’s total reach measure different things. They should not be added together or presented as the number of people using this preset. Broad tags may include older or non-AI content. Use the stated cutoff when quoting a number, and check the original platform for a current reading.

Is the generated scene a record of a real event or evidence about the person pictured?

The video is a staged AI scene based on the supplied photo and the preset’s visual plan. It does not document a real performance, meeting or trip, and it cannot establish a person’s beliefs or hidden characteristics. This preset creates a live-action-style adult performance. It does not turn the person into an anime character, add the anime GIF or supply the Doja Cat recording. The animation reference belongs to the format’s history, while your output uses your uploaded identity. Use a caption that explains the creative scene when context could otherwise be misleading. Historical creator posts in the timeline concern the format’s circulation; they are not evidence that the uploaded person attended an event or participated in the source video.

Where can I see pricing, and what is included when I generate this template?

The pricing page explains the available paid plans and packs. This preset creates one 13-second vertical 9:16 video from one photo, with the scene and role structure described above. Sign in before starting so the task can be associated with your account. Read the purchase and generation information shown for your account before confirming a paid action. Previewing the example or reading the guide does not create a task. A new creative attempt is a separate generation, so review your photos and role order before beginning.

How long should I wait, and where can I find a video after leaving this page?

Generation can take a few minutes and may take longer when the processing queue is busy. Follow the task state in the generator rather than using the preview clip’s length as a waiting-time estimate. The output itself is 13 seconds; processing time is a different measurement. While signed in, check My videos to find the task again. Do not repeatedly submit the same request because the first one is still processing. If the interface reports a failure, use that reported state rather than assuming a temporarily slow task has failed.

Which caption, hashtags and final edits fit this video before I publish it?

One caption idea is “Trying to act normal for six seconds.” Keep enough context to explain what the viewer is seeing without retelling every second. Use the complete trend name or an alias such as Talk to Me Doja AI dance, then add only tags that actually describe the visible scene. Add readable words after downloading instead of expecting the generator to render perfect text. Check the face, transition, ending and audio at normal playback speed. Use the destination platform’s AI disclosure tools where required and confirm that any added music is available for your intended use.

How should I choose a related template if this movement does not match my idea?

Choose by the action, accepted subjects and role count before comparing the title alone. This scene uses one photo and moves from “Lean toward the lens with a composed expression” to “Complete the sway and release the hands”. A dance duet needs different references from a solo portrait, a pet reaction or a multi-shot road scene. The six related cards link to distinct pages rather than alternate spellings of this one. Read the photo requirements on the next page, particularly when moving from one subject to two, and confirm the new scene before uploading again.

Other AI video templates