1. Workflow Overview

• Purpose: Generates portrait posters with a retro Kodak film aesthetic, supporting both online and local execution modes.

• Core Models:

  • Main Model: 高级质感人像-FLUX_v1.0 (realistic portraits).

  • LoRA Adapters:

    • miluo丨胶卷人像_柯达风_东亚定制版-D2.02 (Kodak film tones).

    • F1.0 胶卷质感写实人像摄影—小说推文_V1.0 (enhances facial details).

2. Key Nodes

• CheckpointLoaderSimple: Loads the main model (e.g., 高级质感人像-FLUX_v1.0).

• LoraLoader: Applies Kodak-style LoRA with strength 1.

• CLIPTextEncode: Processes prompts (e.g., miluokd,film style to trigger film effects).

• KSampler: Uses euler sampler with 30 steps.

• VAEDecode: Decodes latent images to final output.

• Special Nodes:

  • BaiduTranslateNode (manual install): Translates prompts (CN↔EN).

  • ttN text/concat: Requires ttN nodes package for dynamic prompt concatenation.

3. Workflow Structure

• Group 1: Online Mode

  • Input: Predefined prompts (e.g., A young East Asian woman...).

  • Output: Quick film-style portraits.

• Group 2: Local Mode

  • Input: Custom prompts via ttN text.

  • Output: High-res posters (768x1024) with finer control.

• Group 3: Prompt Processing

  • Uses ttN concat to merge fixed style keywords (miluokd,film style) with user input.

4. Input & Output

• Inputs:

  • Resolution: 768x1248 (online) or 768x1024 (local).

  • Prompts: Must include triggers like miluokd.

  • Seed: Fixed or randomized.

• Output: PNG-format vintage portrait posters.

5. Notes

1. Model Dependencies:

   • Download main model: LibLib.art Link

   • Download LoRA: Kodak LoRA

2. Hardware:

   • Recommended 8GB+ VRAM; local mode demands higher resources.

3. Troubleshooting:

   • If BaiduTranslateNode is missing, install pysssss via ComfyUI Manager.