Unverified50% confidenceFactExact time
By default, EmbeddingGemma 2 uses 280 tokens per image, 140 tokens per video frame, approximately 25 tokens per second of audio, and has a context window of 8,192 tokens.
1
Sources
50%
Confidence
Long-term
Relevance
10/7/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe model contains approximately 20.8 million parameters, with channel counts progressing as 96→192→384, and supports generating 16-frame video clips at a resolution of 64×64.72% similarUnverifiedDuring text-to-video generation, the system performs video dimension alignment—for example, aligning 400×400 to a multiple of 32 to meet the input requirements of the generation model.71% similarUnverifiedIn video generation, the computational cost of self-attention is proportional to the square of the number of tokens; after VAE encoding and patching, a 4-second 720p 24fps video can yield tens of thousands to over a hundred thousand tokens in total.70% similarVerifiedA typical EnCodec configuration uses 8 layers of RVQ with a codebook size of 1024 per layer, producing 75 frames per second at a 24kHz sampling rate, compressing 1 second of audio into 600 tokens.70% similarUnverifiedThe LTX video generation model has 22 billion parameters yet occupies only 14 GB of memory, taking approximately 13 seconds to generate one second of video.69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/984899API
curl https://kongchang.com/api/v1/knowledge/claims/984899MCP
get_claim(id=984899)