Note: This should work for any OpenAI-compatible API server.

STEP 1: Start an OpenAI-compatible API server, for example you can use these params on Runpod:
IMAGE: openmmlab/lmdeploy:v0.6.2.post1-cu12
COMMAND: bash -c "pip install timm==1.0.7 && lmdeploy serve api_server OpenGVLab/InternVL2-40B-AWQ --model-name OpenGVLab/InternVL2-40B-AWQ --backend turbomind --server-port 3000 --model-format awq --quant-policy 4"
VOLUME PATH: /root
CONTAINER DISK: 10GB
VOLUME DISK: 60GB
PORT: 3000

STEP 2: Enter your server endpoint url in the input below (e.g. something like like 'https://uf1kmzigq0p5bd-3000.proxy.runpod.net' if you're using the above server command).

STEP 3: Enter a system prompt and a prompt at the top of one or more columns, and optionally a prefix that you want the resulting caption to start with.

STEP 4: Select some images to test your prompt on (Ctrl+Click to select multiple images).

STEP 5: Click 'Compute captions for all columns' or click one of the individual column buttons.