Yes. With Face Swap enabled, upload a reference photo and specify who to swap in the prompt. For example: “Replace only the face of the person on the right wearing the blue jacket. Keep everyone else’s face unchanged.”
If they move around, include a detail like their clothing so the model can identify the same person throughout the clip. It’s prompt-based, so accuracy can vary if people cross paths or the target’s face is obscured.

Replies · 36
@gitcon2026MiniMax H3 Studio: httpcore.ConnectError: All connection attempts failed App: MiniMax H3 Studio (minimax-h3-studio.git) Repo: https://github.com/erdinoral/minimax-h3-studio.git Generated: 2026-11-14T07:32:17.853Z Pinokio: 8.3.7 Platform: win32 x64 Node: v24.22.1 Summary System { "pinokio": { "version": "8.3.7",
Thanks for sharing the logs! ComfyUI crashes on startup in comfy-kitchen’s conv3d.py with “NameError: name 'List' is not defined.” H3 Studio then reports a connection error because ComfyUI never starts.
Please stop H3 Studio in Pinokio, run Update, then start it again. The ComfyUI build I checked pins comfy-kitchen 0.2.41, which uses list[int] at that location. If the same error persists, please share the comfy-kitchen version installed in the app’s Python environment and the first traceback from the new ComfyUI startup log. There’s no need to reset the app or re-download your models for this check.
The fix is out! Please Stop the app in Pinokio and Start it again (it updates itself on start). If you prefer, you can also click Update first. After that, SageAttention will be switched off automatically on your RTX 5060 Ti, because the Linux SageAttention package doesn't support RTX 50-series cards yet. Renders will be a bit slower, but they should work. Try a short 5 s video first and let me know how it goes. If it still fails, send me the full error line.
Thanks for confirming! Linux uses a different install path than Windows, so a few quick checks:
Click Update in Pinokio, then restart the app. The error message will now show the real cause.
Turn Spectrum OFF and try a 5 s video. Does it work?
Send me the output of nvidia-smi (we need the driver version; 580+ is required).
If it still fails, send the full error line, or the file ~/pinokio/api/minimax-h3-studio.git/logs/api/start.js/latest.
¡Hola! Gracias por probar MiniMax H3 Studio y por enviar el informe.
Lamentablemente, la aplicación no puede funcionar en tu ordenador. El problema no es la descarga lenta, es el hardware:
Tarjeta gráfica: MiniMax H3 necesita una tarjeta NVIDIA con CUDA. Tu equipo tiene una Intel UHD 610, que es la gráfica integrada del procesador. No tiene CUDA y solo comparte 1 GB de memoria, cuando el modelo necesita al menos 16 GB de VRAM.
Memoria RAM: tienes 8 GB, y ya estaban casi todos ocupados durante la instalación. El modelo es muy grande, unos 60 GB en disco, y necesita mucha más memoria.
Procesador: el Pentium G5400 tiene 2 núcleos. Incluso en modo CPU, un vídeo tardaría muchísimo o simplemente fallaría.
Aunque la descarga terminara, la instalación descargaría decenas de GB y el primer vídeo fallaría.
Qué te recomiendo:
Detén la instalación en Pinokio.
Borra la carpeta C:\pinokio\api\minimax-h3-studio.git para recuperar el espacio.
Para usar esta aplicación hace falta un PC con una NVIDIA de 16 GB de VRAM o más (por ejemplo, RTX 4060 Ti 16 GB, RTX 5060 Ti 16 GB o RTX 4090) y, idealmente, 32 GB de RAM.
En la próxima versión, el instalador comprobará la tarjeta gráfica al principio y avisará antes de descargar nada, para que no vuelva a pasar. ¡Perdona las molestias!
I will soon publish a promotional video and a "how-to" guide on YouTube, and I’ll share them here as well. I believe this will result in content that is both useful and entertaining. I haven’t seen anyone else do this before; if we can pull it off without any hitches, it will be a highly effective option.
Hi! Could you share a screenshot of the exact error and the last 30–50 lines from the H3 Studio terminal in Pinokio, including any traceback? Please remove any API keys or personal information before posting. Also, which OS, GPU/VRAM and Pinokio version are you using, and is this a fresh install or an updated one? Does the Studio page open at all, and does the error happen during startup or only when you use Director? If Studio opens, please tell me which LLM provider/model is selected in Settings and whether Ollama is running if you use it. This will help me identify the cause rather than guess. Thanks!
@cocktailpeanutGenerated with Minimax H3 Singularity (4 steps), NVIDIA A4500. Prompt: 15-second anime video. Use the provided image as the exact opening frame. Preserve the original anime style, character design, red shirt, brown shorts, and proportions throughout. One continuous shot. 0–7s: The muscular man carries an enormous nucl
@cocktailpeanutNo it's in turkish. The launcher menu items are all in Turkish, and the main web ui right after launching is also in Turkish. I am actually running this on my machine right now, not speculating. My machine's locale is english so I'm pretty sure the app is explicitly setting it to Turkish somewhere
Hello! Currently, the application supports Turkish and English. For video production, I have launched with support for four languages, and I plan to increase the number of languages available for both the user interface and the production process.
Regarding the production side: to facilitate the sourcing of visual assets, the platform integrates with the Pexels and Pixabay APIs, allowing for free use—users simply need to create a free account and obtain an API key. Once H3 Studio is installed and active, videos can be generated via a bridge connection. Tasks such as subtitle generation and video assembly are handled locally.
@franckd971MiniMax H3 Studio: INFO: 127.0.0.1:63054 - "GET /api/studio/settings HTTP/1.1" 404 Not Found App: MiniMax H3 Studio (minimax-h3-studio.git) Repo: https://github.com/erdinoral/minimax-h3-studio.git Generated: 2026-10-21T11:45:30.689Z Pinokio: 8.2.0 Platform: win32 x64 Node: v22.21.1 Summary System { "pinokio": { "version": "8.2.0",
Thanks for the report. We traced this 404 to a leftover request from the frontend for /api/studio/settings; the current server does not expose that endpoint, and the UI falls back to the setting being off. This log alone should not stop video generation. Did you notice any other visible problem besides this 404?
Hi! Start in the Scene tab for your first video:
1. Choose New Video.
2. Describe what should happen, including the subject, action and camera movement.
3. Start with a short 5-second draft to test your prompt.
4. Click Generate. Once finished, you can preview and download it, or choose Continue from this video.
For a longer story with multiple scenes, use the Director tab. You can create character and location references, add scene cards, and generate them in sequence.
What kind of video would you like to make? I can help you set up your first prompt.
