Using GPU on windows 11 with Immich for Encoding and Decoding.

[Scopes1940]

Attached is my process using chat gpt as a recap to fix any issues etc been having a hard time as a person who does not use linux or any type of code but i love the idea of Immich!
PS: you may need to install Ubuntu 22.04.5 LTS or later
Immich FFmpeg.docx

In case you acnnot access the word document:

✅ Immich FFmpeg & GPU Acceleration Setup Recap

1. Verify Containers Are Running
   Check your Docker containers to make sure Immich services are up:
   docker ps
   You should see:
   • immich_server
   • immich_postgres
   • immich_machine_learning
   • immich_redis

---

2. FFmpeg Binary
   • Immich uses FFmpeg to generate thumbnails and transcode videos.
   • Inside the server container, check FFmpeg is installed:
   docker exec -it immich_server /usr/bin/ffmpeg -version
   • The container path /usr/bin/ffmpeg works.
   • Do not try to symlink to /usr/local/bin/ffmpeg, the directory is read-only in the container.

---

3. Setting FFMPEG_PATH
   • In your .env file, set:
   FFMPEG_PATH=/usr/bin/ffmpeg
   • Make sure the container picks this up:
   docker exec -it immich_server printenv | findstr FFMPEG_PATH
   • This avoids EACCES errors from trying to spawn the wrong path.

---

4. Docker Compose Volumes
   • Only mount the real FFmpeg binary if needed, otherwise let the container use its internal one:
   volumes:

• ${UPLOAD_LOCATION}:/data
• /etc/localtime:/etc/localtime:ro
  • Avoid mounting to /usr/local/bin/ffmpeg unless you really need a custom version.

---

5. Upload Permissions
   • Uploaded media folder should have full read/write permissions for the container. On Windows:
   icacls D:\Immich /grant "Everyone:(OI)(CI)F"
   • This avoids errors like:
   Key ("ownerId", checksum) already exists
   • Don’t try to “make folders read/write inside the container” — fix permissions on the host.

---

6. GPU Acceleration
   • Immich supports NVENC (NVIDIA) for hardware encoding.
   • For end-to-end acceleration (decode + encode), enable:
   Hardware decoding
   in the web interface settings → transcoding.
   • Your GPU graph will spike when decoding videos — this is expected. CPU usage should drop accordingly.
   • In .env / docker-compose, make sure:
   o Container uses runtime: nvidia
   o Docker has access to the GPU

---

7. Don’t Do These
   • Don’t try to chmod +x /usr/local/bin/ffmpeg — container filesystem is read-only.
   • Don’t mount FFmpeg over /usr/local/bin/ffmpeg unnecessarily.
   • Don’t change database container folders to network shares — Immich database must be local.
   • Don’t delete or reset .env mid-run — it can break connections and container environment.

---

8. Verifying It Works
9. Upload a video (like 4K 60fps) and watch the logs:
   Transcoding video ... with NVENC-accelerated encoding
10. Watch GPU utilization graph:
    o Spikes indicate hardware decoding is active.
11. CPU usage should drop slightly.
12. FFmpeg errors (EACCES) should disappear after FFMPEG_PATH=/usr/bin/ffmpeg.

---

This setup ensures:
• Immich uses its internal FFmpeg binary safely.
• GPU acceleration works for encoding and decoding (if enabled in settings).
• Uploaded media is fully accessible.
• No more permission errors.

---

I can also make a shorter “copy-paste checklist” version if you want to share with someone quickly.
Do you want me to do that?
You said:
yes include that too
ChatGPT said:
Perfect! Here’s a concise copy-paste checklist version you can share:

---

📝 Immich FFmpeg + GPU Acceleration Checklist

1. Verify Containers
   docker ps
   Check for:
   • immich_server
   • immich_postgres
   • immich_machine_learning
   • immich_redis

---

2. FFmpeg
   docker exec -it immich_server /usr/bin/ffmpeg -version
   • Internal FFmpeg works.
   • Do not symlink or chmod /usr/local/bin/ffmpeg (read-only).

---

3. .env Settings
   FFMPEG_PATH=/usr/bin/ffmpeg
   UPLOAD_LOCATION=D:\Immich
   DB_DATA_LOCATION=./postgres
   IMMICH_VERSION=release
   DB_PASSWORD=Your own password

---

4. Volumes (docker-compose)
   volumes:

• ${UPLOAD_LOCATION}:/data
• /etc/localtime:/etc/localtime:ro
  • Avoid mounting over /usr/local/bin/ffmpeg.

---

5. Folder Permissions
   • Make uploads folder fully readable/writable:
   icacls D:\Immich /grant "Everyone:(OI)(CI)F"

---

6. Enable GPU Acceleration
   • In web UI: Settings → Transcoding → Hardware decoding.
   • Ensure Docker uses NVIDIA runtime:
   runtime: nvidia
   • Check GPU graph for spikes during video transcoding.

---

7. Do Not
   • Don’t chmod/mount /usr/local/bin/ffmpeg.
   • Don’t use network shares for database.
   • Don’t reset .env mid-run.

---

8. Verify It Works
   • Upload video → check logs for:
   Transcoding video ... with NVENC-accelerated encoding
   • No EACCES errors.
   • CPU usage drops, GPU spikes.

---

This ensures Immich works safely with GPU acceleration and avoids FFmpeg permission errors.

PS: if you plan to use an AI to help with this process as i did give it a compy paste version fo your .yml and .env files
TAKE EACH STEP ONE AT A TIME IF YOU ARE A BEGINNER
Fianlly for the Decode to work ensure you fix the hwaccel.yml first then enable it in the settings in Immich UI as thats what worked for me.