Back to blog

7 Photo Tips for Better Talk to Me Doja AI Dance Videos (Lighting, Hands & Framing)

Nov 17, 2026Talk to Me Doja Editorial TeamTalk to Me Doja Editorial Team

Follow these 7 Talk to Me Doja photo tips to get clean cheek-cupping hands, sharp facial expressions, and zero safety rejections on your 12s 9:16 AI video.

Start with a well-lit, front-facing photo that leaves the cheeks and chin visible. The Talk to Me Doja AI Video Generator uses it as the reference for a 12-second clip with a cheek cup and side-to-side sway.

Below are the seven photographic rules we tested across hundreds of bytedance/seedance-2-mini (Fast) and bytedance/seedance-2-fast (Premium) renders, along with a diagnostic table comparing ideal inputs against common mistakes.

Quick Diagnostic Table: Good vs. Risky Source Photos

Before uploading to the Talk to Me Doja AI Studio, compare your candidate photo against these technical specifications:

Photo AttributeIdeal Input (Best Result)Risky Input (Avoid)Impact on 12s Animation
Framing & Crop

Waist-up or chest-up portrait with shoulders visible

Tight chin-only crop or extreme macro

Visible sleeves help the model raise both hands cleanly at Beat 2 (2–5s).

Head AngleFront-facing (0°–15°) with both eyes openSide profile or looking down at the floor

Symmetric facial landmarks keep the 5–8s bashful sway stable.

LightingEven daylight or diffused indoor lighting

Harsh split shadows, dark rooms, or strong backlight

Prevents shadow flicker across the cheeks during the head tilt.

Hand PositionArms relaxed at sides or hands unoccluded

Holding a coffee cup, phone, or touching glasses

Eliminates ghost props sticking to the palms during the cheek cup.

Resolution & File

JPG, PNG, or WebP (800×1200 px+, under 20 MB)

Below 300×300 px or blurry screenshots

Gives Premium 720p (45 credits) enough detail for crisp eyes and fingers.

Subject RulesOne consenting adult in a personal photo

Celebrities, minors, multiple people, or anime stills

Public figures and anime frames are blocked (credits auto-refunded 100%).

The 7 Essential Talk to Me Doja Photo Tips

1. Frame From the Chest or Waist Up (Give the Arms Room to Move)

In the original Episode 1 scene of Yamada-kun and the Seven Witches, Urara Shiraishi is framed from the upper torso up so the viewer can see her elbows bend and her palms press against her cheeks. If you upload an ID-style crop that cuts off right below the chin, the video model has zero reference for your shirt sleeves, skin tone on the wrists, or shoulder width. Choosing a waist-up or half-body portrait gives Seedance 2.0 the anatomical context it needs to bring both hands up smoothly at the 2-second mark.

2. Keep Props Away From Your Chin and Hands

Mirror selfies where you hold an iPhone directly in front of your collarbone are the number-one cause of distorted hands. When the model transitions from Beat 1 (0–2s Cool Pose) into Beat 2 (2–5s Cheek Cup), it tries to move your hands toward your jawline—and if a phone or iced coffee is glued to your fingers in frame zero, the object stretches along with your palm. Set your phone on a timer or crop out any held items before uploading.

3. Face the Camera Lens Directly (Within 15 Degrees)

During Beat 4 (8–10s Shy Glance) and Beat 5 (10–12s Camera Lean), the choreography relies on looking away briefly and then re-locking eye contact with a bashful smile. When your source photo is a three-quarter or 90-degree side profile, the model has to hallucinate the hidden side of your face as you sway left to right. A straight-on portrait where both ears and eyes are visible keeps your likeness locked from start to finish.

4. Use Flat, Front-Facing Lighting Instead of Dramatic Shadows

Moody bedroom lighting or golden-hour side shadows look great in still photography, but they behave like painted-on textures in single-photo video generation. As your head tilts left and right between seconds 5 and 8, a dark shadow cast by your nose can appear to slide across your cheek. Stand facing a window, an overcast sky, or a balanced indoor lamp so both cheeks have matching brightness.

5. Watch Out for Thick Eyeglass Frames and Heavy Bangs

If sunglasses, thick frames, or bangs hide your eyes, try a photo where both eyes are visible. When you review the video, look at the glasses and eyes as the person glances away and back.

6. Show the Outfit You Want the Model to Reference

Use a photo that clearly shows the outfit you want the model to reference. Try simple colors if small lettering or dense patterns are hard to make out at 480p.

7. Never Upload Celebrities, Minors, or Anime Stills

To keep our service safe and reliable, every upload goes through three automatic checks:

  1. Instant browser and server validation for file format (JPG, PNG, WebP), size (≤ 20 MB), and dimensions (≥ 300 × 300 px).
  2. Mandatory confirmation that you have permission from the person pictured.
  3. Upstream Kie / Seedance safety moderation, which automatically rejects celebrities, public figures, minors, NSFW content, and copyrighted anime screenshots.

If your task is marked failed or rejected after credits were deducted, the credits used for that task automatically return to your balance.

Which Model Tier Should You Pair With Your Photo?

Once you have a clean half-body portrait, you can match it to the right credit tier on our pricing page:

  • Well-lit, simple background portrait: Start with Fast · 480p (15 credits, $4.99 in Starter) or Fast · 720p (22 credits).
  • Use Seedance 2.0 Fast: Choose Premium 480p (30 credits) or Premium 720p (45 credits).

Anyone with a working media link can open the file, so share links carefully. When your photo is ready, follow the step-by-step tutorial or go to Studio.