Skip to content

TUTORIAL_EN.md

Latest commit

 

History

History
93 lines (67 loc) · 3.51 KB

TUTORIAL_EN.md

File metadata and controls

93 lines (67 loc) · 3.51 KB
Raw

Simple Video Tutorial

This tutorial provides the shortest path to try each feature.

1. Start

cd simple_video_app
./start.sh

Browser: http://127.0.0.1:8090/

2. Create a Video (T2V)

  1. In ⚙️ Generation Sequence, select a T2V preset
  2. In ⚙️ Video Settings, adjust scene count and scene length
  3. In 📜 Scenario Input, write what you want to generate
  4. If needed, click 🎨 Style Preset to inject style hints (default scene length becomes 5s when applied)
  5. Build scene prompts in two steps: 🧠 Build Scenario → 🤖 Generate Prompts
  6. Click Generate Video
  7. Check results in 🎬 Output

3. Create Character Video (I2V / FLF)

  1. Add reference images to 📥 Image Drop (ref1–ref3)
  2. Select from 👤 Character List or register new via 📝 Register Character
  3. Enter initial image prompt in 📝 What do you want to draw?
  4. Click Generate Initial Image (auto-reflected as key image)
  5. In ⚙️ Generation Sequence, select a character preset (I2V / FLF)
  6. Click Generate Video

Notes:

  • In char_edit_i2i_flf, initial image and character image are especially important
  • If intermediate confirmation is enabled, click CONTINUE to proceed

4. Create a Mixed Transition Video (I2I + Auto FLF/I2V)

With Key Image Ref I2I + Mixed Transitions or Character EDIT + I2I + Mixed Transitions, the LLM judges the relationship between scenes and automatically assigns FLF (smooth morph) or I2V (motion-first) per boundary.

Key Image variant (no character image needed)

  1. Upload a reference image to 🖼️ Key Image
  2. In ⚙️ Generation Sequence, select 🔀 Key Image Ref I2I + Mixed Transitions
  3. In ⚙️ Video Settings, set scene count (internally N+1 images are generated)
  4. Enter a scenario and run 🤖 Generate Prompts
    • After generation, a 🔀 Scene Transitions editor appears below the prompts
    • Optionally change any boundary type (FLF / Cut / Crossfade / Fade Black)
  5. Click 🖼️ Pre-generate Scene Images to generate N+1 scene still images
    • Images can be reviewed and replaced individually
  6. Click CONTINUE to proceed to the video generation phase (FLF or I2V runs per boundary)
  7. Final concatenated video appears in 🎬 Output

Character EDIT variant

Same steps, but requires a 🖼️ Key Image (character reference) and character selection/registration.
The EDIT → I2I two-stage pipeline produces higher character consistency.

Notes:

  • The 🔀 Scene Transitions editor is collapsible (accordion); manual changes are saved instantly
  • LTX workflows are not available for these presets (I2V always uses WAN)

5. Create Music (T2A)

  1. Enter mood and lyric ideas in the music input area
  2. Use lyric/tag suggestions if needed
  3. Run music generation and check the output audio

6. Create Video from Music (M2V)

  1. Select generated audio or uploaded audio
  2. Enter a video scenario
    • If empty, a confirmation dialog appears
    • Choose either Generate as-is or Input scenario
  3. Run Generate Video
  4. Final output is saved to output/movie

7. Create Music from Video (V2M)

  1. Select generated video or uploaded video
  2. Set music style conditions if needed
  3. Run and check the merged result
  4. Final output is saved to output/movie

8. Troubleshooting

  • If ComfyUI /prompt failed appears, check where it stopped in error details (node_errors)
  • For reference image errors, re-check 📥 Image Drop and selected character state
  • Old job_id 404 errors are often resolved by regenerating