the design it quite robotish, make the dark them design reddish and black themes with
- User
the design it quite robotish, make the dark them design reddish and black themes with
- Assistant
I'll restyle the entire app into a richer reddish-black dark theme—warmer, less clinical, more cinematic—with crimson accents, soft ember glows, and refined glass surfaces.
The whole UI is now a reddish-black crimson theme instead of the cool indigo/robot look.
Design updates
- Base: near-black
#0a0505with deep red glass panels - Accents: rose / crimson / ember orange gradients
- Glow: warm red ember ambient light and soft vignettes
- Borders & inputs: red-950 / rose borders instead of slate/indigo
- Buttons: red → rose → orange fire gradients
- Brand: flame icon + “CRIMSON” badge
- Surfaces: rounded softer panels, less clinical chrome feel
Functionality is unchanged — chronological rename, caption studio, sidebar list, thank-you page, and folder/ZIP download all still work the same.
Files built:
src/index.css,index.html,src/components/Navbar.tsx,src/components/SettingsModal.tsx,src/components/UploadScreen.tsx,src/components/ReviewScreen.tsx,src/components/CaptioningStudio.tsx,src/components/ThankYouScreen.tsx,src/components/DownloadModal.tsx,src/App.tsx - Base: near-black
- User
perfect add a api key from open router and nvidea nim for the cusom vison models that can see the image and based on that image provides the prompts , also in setting tab or any ther particular section give an instruction prompt for ai visin model how it should describe the image. a button should be there for activating the ai prompting in the captioning section
- Assistant
I'll add OpenRouter + NVIDIA NIM vision captioning: API keys, model pickers, a custom instruction prompt in Settings, and an AI Prompt button in the caption studio.
AI vision captioning is wired in for OpenRouter and NVIDIA NIM.
What’s new
Settings → AI Vision Models
- Choose provider: OpenRouter or NVIDIA NIM
- Paste API keys (
sk-or-v1-.../nvapi-...) - Pick vision models (or paste a custom model id)
- Instruction prompt for how the model should describe images (editable + reset to default)
- Temperature, max tokens, auto prefix/suffix on AI captions
- Keys & settings saved in
localStorage(browser only)
Caption Studio
- AI Prompt from Image button — sends the current image + your instruction prompt to the selected vision model
- Loading overlay while the model “sees” the image
- Generated caption fills the description box (editable before Next)
- Keys shortcut opens settings if the API key is missing
How to use
- Open Settings
- Add OpenRouter and/or NVIDIA key
- Set provider + model + instruction prompt
- In the studio, click AI Prompt from Image
- Edit if needed → Next
Images are resized/compressed client-side before upload to keep API payloads light.
Files built:
src/types/dataset.ts,src/utils/visionApi.ts,src/components/SettingsModal.tsx,src/components/CaptioningStudio.tsx,src/App.tsx
Other models, same or similar prompt
Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.