cstr/index-echo-9b-GGUF overview
Index Echo S2TT 9B — CrispASR GGUF Native speech translation with bilingual, timestamped subtitles, primarily Chinese → English, Japanese or Spanish. This repo…
Runs locally from ~1.22 GB disk (4 GB VRAM class GPUs with llama.cpp / guIDE).
Repository Files & Downloads
| File | Type | Quantization | Size | Link |
|---|---|---|---|---|
| experiments/q4-guards-20261008/q4_k_ffn_guarded/decoder.gguf | GGUF | GGUF | 10.97 GB | Download |
| experiments/q4-guards-20261008/q4_k_middle/decoder.gguf | GGUF | GGUF | 13.46 GB | Download |
| experiments/q4-guards-20261008/q4_k_plain/decoder.gguf | GGUF | GGUF | 4.70 GB | Download |
| experiments/q4-guards-20261008/q4_k_sensitive/decoder.gguf | GGUF | GGUF | 9.51 GB | Download |
| experiments/q4-imatrix-20261009/q4_k_ffn_guarded/decoder.gguf | GGUF | GGUF | 10.97 GB | Download |
| experiments/q4-imatrix-20261009/q4_k_middle/decoder.gguf | GGUF | GGUF | 13.46 GB | Download |
| experiments/q4-imatrix-20261009/q4_k_plain/decoder.gguf | GGUF | GGUF | 4.70 GB | Download |
| index-echo-9b-decoder-f16.gguf | GGUF | F16 | 16.69 GB | Download |
| index-echo-9b-f16.gguf | GGUF | F16 | 1.22 GB | Download |
Model Details
Model README
---
license: apache-2.0
base_model: IndexTeam/Index-Echo-S2TT-9B
tags:
- gguf
- speech-translation
- crispasr
---
Index-Echo S2TT 9B — CrispASR GGUF
Native speech translation with bilingual, timestamped subtitles, primarily
Chinese → English, Japanese or Spanish. This repository contains the validated
F16 encoder/connector and decoder pair for CrispASR.
| File | Bytes | GiB |
|---|---:|---:|
| index-echo-9b-f16.gguf | 1,313,848,736 | 1.224 |
| index-echo-9b-decoder-f16.gguf | 17,920,696,992 | 16.690 |
| Pair | 19,234,545,728 | 17.914 |
Use
Keep both GGUF files in the same directory. Use a CrispASR build containing
earlier 2B-only builds cannot load this pair.
crispasr -m index-echo-9b-f16.gguf --auto-download -f speech.wav --target-lang en -osrt--target-lang selects en, ja or es. Complete-file processing also uses
the standard ggml-silero-v6.2.0.bin VAD companion; --auto-download fetches it.
The 2B model remains CrispASR's smaller default. This 9B repository supports F16 only.
Memory and validation
Allow about 18 GiB for weights plus working memory. CPU inference and CUDA
inference were validated; the CUDA runs used two 15 GiB Tesla T4 cards.
Host RAM requirements also depend on model loading and mmap behavior.
The F16 pair passes focused original-reference stage, decoded-output and
timestamp controls on CPU and CUDA, including speech roundtrips. Immutable
pins, hashes and evidence are in acceptance.json.
Detailed results, rejected experiments and profiling measurements are archived
Provenance and license
Source: IndexTeam/Index-Echo-S2TT-9B,
revision b8ac6fb7d3dc17cee48a52201bd3d93dc86b0dba.
Decoder conversion: ggml-org/llama.cpp@42d958167a748f2c04b1f888e84e7a58f609ddcb,
F16 with --no-mtp.
Apache-2.0; LICENSE is copied from the pinned
Run cstr/index-echo-9b-GGUF with guIDE
Download guIDE — the AI-native code editor with local LLM inference and 69 built-in tools.
Source: Hugging Face · Compare models