TinyPine Studio runs on Windows with an NVIDIA graphics card. 24 GB of graphics memory is recommended, because video needs it. 8 GB renders pictures at smaller sizes, and 16 GB runs everything except video. The app reads your GPU and sets sizes, precision and a memory ceiling to match. Pick a card size.
Measured on disk with everything installed. Start with the four marked rows, about 43 GB; every other part downloads from the Grove the first time you use a feature that needs it, after asking. With every part installed the library holds about 350 GB of models. The library can live on any drive and move later.
| Part | On disk | What it is for |
|---|---|---|
| App enginestart | 5.6 GB | Python and the CUDA libraries. Installed on first start. |
| Krea 2 Turbo, Qualitystart | 33.2 GB | Picture generation. Every picture needs it. |
| Assist 4Bstart | 2.7 GB | Enrich, Vary, chat and captions. Stored in 4-bit. |
| Depth, upscaler, Facestart | 1.0 GB | Depth effects and focus, ×2 upscale, face landmarks and Face Enhance. |
| Krea 2 Fast (fp8) | +11.9 GB | Fast mode, on 16 GB cards and up. |
| Krea 2 Raw | +24.5 GB | Training your own LoRAs. |
| Assist 8B | 6.0 GB | A stronger Assist, stored in 4-bit. |
| Video: LTX-2.5 and its engine | 79.9 GB | Video with sound and Extend. Also needs a 24 GB card, 32 GB of RAM and a 64 GB page file, which the installer sets up. |
| Upscale video: RelayVSR 2x | 10.3 GB | Videos twice the size, and the ×2 step after a video. About 16 GB of graphics memory. |
| Music: MiniMax-Music3 and its engine | 31.1 GB | Songs and song extension. |
| Speech: MOSS-TTS and its engine | 22.6 GB | Voice-overs in 31 languages. About 13 GB of graphics memory while it speaks. |
| Voice design: MOSS-VoiceGenerator | +11.3 GB | A voice described in words, for any of the 31 languages. |
| English speech: Chatterbox and its engine | 9.0 GB | Livelier English voices. About 4 GB of graphics memory. |
| Sound effects: EzAudio and its engine | 10.8 GB | Sound effects from a sentence. |
| Design generation and layers | 93.4 GB | Generate in Design and Split into layers. The canvas itself needs none of it. |
| Transcription: Whisper | 0.5 to 2.9 GB | Subtitles and transcripts. Small, Turbo or the Swedish model. |
| Selection, cut-out, slow motion, tracking | 1.4 GB | SAM 2.1, Lucida, RIFE and FILM, BootsTAPIR. |