Add video reverse-prompt support and keep original media

2025-12-13 00:14:29 +08:00
parent 4cd29f007e
commit 26e18d1a76
2 changed files with 97 additions and 15 deletions
@@ -7,6 +7,10 @@ Through triples, any LLM model compatible with the OpenAI interface can be acces
 - **apikey**
 - **model**

+## Update: Video reverse-prompt (video captioning)
+- **Optional input `video`**: supports OpenAI-compatible `chat.completions` payload with `{"type":"video_url"}`.
+- **No resize / no compression**: image keeps original resolution; video is read as-is and base64 encoded.
+
 **Online Demo**
 https://www.runninghub.ai/post/1890402871119368194
 ![image](https://github.com/user-attachments/assets/31b35db4-4d61-4767-a41c-6f1445fbea5e)