QuerryRaspBerry commited on
Commit
3920666
·
verified ·
1 Parent(s): 364d09d

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +10 -32
README.md CHANGED
@@ -1,44 +1,22 @@
1
  ---
2
  title: IRIS
3
  emoji: 🔎
4
- colorFrom: indigo
5
- colorTo: pink
6
  sdk: gradio
7
  sdk_version: 6.14.0
8
- python_version: '3.11'
9
  app_file: app.py
10
  pinned: false
11
  license: apache-2.0
12
- short_description: Intelligent search for spoken moments inside video.
13
- suggested_hardware: cpu-upgrade
14
- startup_duration_timeout: 1h
15
- models:
16
- - openai/whisper-tiny
17
- - sentence-transformers/all-MiniLM-L6-v2
18
- tags:
19
- - video-search
20
- - speech-recognition
21
- - semantic-search
22
- - gradio
23
- preload_from_hub:
24
- - openai/whisper-tiny
25
- - sentence-transformers/all-MiniLM-L6-v2
26
  ---
27
 
28
- # IRIS
29
 
30
- IRIS is a Gradio demo for searching videos by spoken content. The Space mirrors
31
- the `iris/` package from `../IRIS-core` so Hugging Face can build it without
32
- cloning a private dependency during installation.
33
 
34
- Upload a video, process it once, then run as many searches as needed against the
35
- processed transcript. The Space defaults to CPU-friendly settings:
36
-
37
- - `IRIS_ASR_MODEL=openai/whisper-tiny`
38
- - `IRIS_DEVICE=cpu`
39
- - `IRIS_MAX_AUDIO_SECONDS=120`
40
- - `IRIS_VISUAL_SEARCH_ENABLED=false`
41
- - `IRIS_MAX_STORED_VIDEOS=3`
42
-
43
- Set Space variables to use larger models, GPU hardware, longer videos, or
44
- visual search.
 
1
  ---
2
  title: IRIS
3
  emoji: 🔎
4
+ colorFrom: blue
5
+ colorTo: indigo
6
  sdk: gradio
7
  sdk_version: 6.14.0
8
+ python_version: '3.13'
9
  app_file: app.py
10
  pinned: false
11
  license: apache-2.0
12
+ short_description: Find the exact moment a topic is mentioned
 
 
 
 
 
 
 
 
 
 
 
 
 
13
  ---
14
 
15
+ # IRIS — Intelligent Retrieval & Indexing System
16
 
17
+ Upload a video + a query; IRIS returns the exact timestamped moments where the
18
+ topic is mentioned or shown (speech via Whisper + vision via CLIP, semantic
19
+ retrieval + cross-encoder rerank). No LLM summary — just the moments.
20
 
21
+ Used through https://app.nnlabs.pl/iris. Requires a signed usage token from the
22
+ website (`NINA_SIGNING_SECRET`); direct API access is disabled.