Samsar Studio Client is the React + Vite frontend for Samsar's image, video, and audio creation workflows.

It currently ships three connected creation surfaces:

• Studio for multi-scene video editing, AI generation, audio, animation, and publishing.

• Image Studio for standalone image generation, prompt editing, inpainting, upload/library workflows, and layered export.

• VidGenie for one-shot video creation with Text to Video and Image List to Video modes, with direct handoff back into Studio.

• Hosted app: app.samsar.one

• API and product docs: docs.samsar.one

Legacy /vidgpt/:id and /videogpt/:id routes still resolve to the current VidGenie experience.

Studio is the full editor. The current toolbar layout includes Defaults, Generate Image, Edit Image, Generate Video, Generate Audio, Actions, Select, Animate, Upload/Library, Text, Shape, and Layers. It is the right surface for multi-scene work, AI-assisted iteration, lip sync, synced sound effects, subtitle alignment, and final publish/export.

Image Studio is the dedicated image workspace. It exposes Generate Image, Edit Image, and Upload/Library directly in the toolbar, supports layered canvases, custom canvas sizes, preset aspect ratios, prompt-based edits, masked inpaint edits, and simple or advanced downloads. The standalone image workspace uses the same shared image-generation and image-edit pipelines as Studio, but with a more streamlined editing UI.

VidGenie is the fast one-shot creator:

• T2V lets you pick an image model, a video model, aspect ratio, duration, language, and subtitle behavior from one screen.
• I2V accepts one or more uploaded images plus an optional prompt hint, then runs the image-list-to-video pipeline.
• Both modes can hand the output back into Studio with View in Studio.

The tables below reflect the current client-side model configuration in src/constants/Types.ts and src/constants/ModelPrices.jsx.

These models are part of the current shared client-side image generation pool. Studio exposes model selection directly, Image Studio uses the same generation pipeline in a simplified flow, and VidGenie exposes a curated subset for T2V.

These edit models are part of the current shared client-side image editing pipeline. Studio exposes model selection directly, while Image Studio focuses the UI on the edit flow itself.

Notes:

• GPTIMAGE2EDIT uses the inpaint flow with a mask brush.
• Image Studio also supports canvas preset changes and custom canvas sizes before generation or editing.

Studio exposes the full video model list. VidGenie T2V exposes a curated subset from the same pool.

T2V currently exposes:

• Image models: GPTIMAGE2, NANOBANANA2, SEEDREAM
• Video models: VEO3.1I2V, VEO3.1I2VFAST, SEEDANCEI2V, KLINGIMGTOVID3PRO, RUNWAYML, SORA2PRO
• Aspect ratios: 16:9, 9:16
• Durations: 10, 30, 60, 90, 120, 180 seconds
• Language selector plus optional subtitles
• Voice dictation for prompt entry

I2V currently exposes:

• Multi-image upload plus optional prompt guidance
• Aspect ratio, language, and subtitle controls
• Direct call into the image-list-to-video creator flow
• View in Studio, publish, and download after render

Studio also includes the following non-generation provider groups:

• Speech/TTS: OpenAI, PlayAI, ElevenLabs
• Music: CassetteAI, AudioCraft, Lyria 2, ElevenLabs Music
• Lip sync: HummingBird Lip Sync, Latent Sync, Sync Lip Sync, Kling Lip Sync, Creatify Lip Sync
• Synced sound effects: MMAudio V2, Mirelo AI

1. Copy the local environment file:

   cp .env.example .env

2. Install dependencies:

   yarn install

3. Start the app:

   yarn start

The dev server is intentionally pinned to port 3000, and VITE_CLIENT_URL in .env.example is already set to http://localhost:3000. Keep that port for local Google OAuth flows.

VITE_PROCESSOR_API=https://api.samsar.one
VITE_CLIENT_URL=http://localhost:3000
VITE_CURRENT_ENV=production
VITE_STATIC_CDN_URL=https://static.samsar.one

yarn build

• Discord: discord.gg/2tbhKwRy
• Gallery: youtube.com/@samsar_one
• X: x.com/samsar_one
• Threads: threads.net/@samsar_one_videos

Contributions and pull requests are welcome.