Lip Sync Talking Photo · skill pública de Open Skill
Lip Sync Talking Photo
Lip Sync Talking Photo es una Open Skill pública en Digen. Ábrela para ver qué crea este flujo.
Film
4 0 4
https://d7qlwi7tl2u27.cloudfront.net/skill-covers/prod/4284959/2026-10-16/tutrxbju.webp
# Lip Sync Talking Clip Type the words. The person in the photo says them on camera with matching mouth shapes, natural blinks and a voice that fits them. ## When to use - "Make this photo say …", talking portrait, speaking avatar - Greetings, announcements, short promos, birthday messages from a portrait - A redo with a different script, mood or language Not for: singing music videos with backing tracks, scenes with several speakers, or animating a photo without speech. ## Inputs and gates | Input | Required | Default | Gate | |---|---|---|---| | Photo | Yes | — | One adult, face frontal or three-quarter, mouth visible and not covered by hands or a mic. Ask for another photo otherwise | | Script | Yes | — | Any language. Fit it to the length (see below) | | Voice mood | No | `warm` | `warm`, `bright`, `calm`, `energetic`, `serious` | | Orientation | No | `9:16` | `16:9` or `1:1` on request | - If the person may be under 18, decline. - No impersonation of real public figures and no deceptive content. Video does not allow NSFW content. **Length fit**: speech runs about 2.5–3 words per second (about 4–5 Chinese characters per second). - 6 s clip: up to about 15 English words or about 25 Chinese characters. - 10 s clip: up to about 25 English words or about 45 Chinese characters. - Longer scripts: offer a tightened version and confirm it before generating. Never let speech get cut off mid-word. ## Steps Read `references/lipsync-direction.md` for the mood table and fixes. 1. **Read the person**: `recognize_image(image_url=<photo>, prompt='Describe this adult: apparent age range, gender presentation, face framing (frontal/3-4), mouth open or closed, expression, setting, lighting. One compact line.')`. If the face is in profile or the mouth is hidden, ask for another photo. 2. **Pick the duration**: 6 s or 10 s, from the script length. 3. **Voice**: `generate_tts(text=<script>, model='tts.design', voice_instruction=<VOICE>, name='Talking clip voice', entity_type='voiceovers')`. 4. **Talking video**: `generate_video(prompt=<PERFORMANCE>, input_urls=[photo], audio_url=<tts audio url>, model='ai2v', orientation=<o>, aspect_ratio=<ar>, duration=<d>, name='Talking Clip', entity_type='final_videos')`. Pass the audio only in `audio_url`, never in `input_urls`. 5. **Self-check, at most one retry**: if the mouth looks mushy or out of sync, retry once with the performance prompt simplified (static camera, tighter framing, explicit "speaks clearly with accurate lip sync"). 6. Reply in one or two sentences and offer another mood. ## Prompt templates **VOICE** (acoustic only; never put the script here) ```text A natural <age range> <gender presentation> voice, <mood voice cues>, clear close-mic studio sound, no background music, native pronunciation in <script language>. ``` **PERFORMANCE**: one flowing paragraph in present tense. The audio drives the mouth; this prompt drives the performance. ```text A medium close-up of the same person from the image speaking directly to camera. Their face, hair, makeup and clothing stay exactly as in the photo. They deliver the words with <mood physical cues>, natural blinks and small natural head movements, shoulders rising slightly with each breath. The background stays as in the photo, softly out of focus. The camera holds a steady frame with a very slow push-in, face sharp, eyes in the upper third. Soft flattering key light consistent with the photo. The person speaks clearly to camera with accurate lip sync; hands stay below the frame. ``` Mood table: | Mood | `<mood voice cues>` | `<mood physical cues>` | |---|---|---| | warm | friendly mid pace, gentle smile in the tone | soft eyes, gentle smile between phrases, relaxed shoulders | | bright | upbeat, lively pitch, slightly quick | lifted brows, quick friendly nods, open smile | | calm | slow even pace, soft volume | slow blinks, minimal head motion, steady gaze | | energetic | projected, fast, punchy emphasis | wide eyes, forward lea