Instructions to use archgod94/tellcue-models with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use archgod94/tellcue-models with llama.cpp:

Install (macOS, Linux)

curl -LsSf https://llama.app/install.sh | sh
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf archgod94/tellcue-models:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf archgod94/tellcue-models:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf archgod94/tellcue-models:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf archgod94/tellcue-models:Q4_K_M

Build from source code

git clone https://github.com/ggml-org/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-app
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama serve -hf archgod94/tellcue-models:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama cli -hf archgod94/tellcue-models:Q4_K_M

Use Docker

docker model run hf.co/archgod94/tellcue-models:Q4_K_M

Llama new
LM Studio
Jan

vLLM

How to use archgod94/tellcue-models with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "archgod94/tellcue-models"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "archgod94/tellcue-models",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/archgod94/tellcue-models:Q4_K_M

Ollama
How to use archgod94/tellcue-models with Ollama:
```
ollama run hf.co/archgod94/tellcue-models:Q4_K_M
```
Unsloth Desktop

How to use archgod94/tellcue-models with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf archgod94/tellcue-models:Q4_K_M

Configure the model in Pi

# Install Pi:
npm install -g @earendil-works/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "archgod94/tellcue-models:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use archgod94/tellcue-models with Docker Model Runner:
```
docker model run hf.co/archgod94/tellcue-models:Q4_K_M
```

Lemonade

How to use archgod94/tellcue-models with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull archgod94/tellcue-models:Q4_K_M

Run and chat with the model

lemonade run user.tellcue-models-Q4_K_M

List all available models

lemonade list

Hermes Agent

How to use archgod94/tellcue-models with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf archgod94/tellcue-models:Q4_K_M

Configure Hermes

# Install Hermes:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
hermes setup
# Point Hermes at the local server:
hermes config set model.provider custom
hermes config set model.base_url http://127.0.0.1:8080/v1
hermes config set model.default archgod94/tellcue-models:Q4_K_M

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use archgod94/tellcue-models with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf archgod94/tellcue-models:Q4_K_M

Configure OpenClaw

# Install OpenClaw:
npm install -g openclaw@latest
# Register the local server and set it as the default model:
openclaw onboard --non-interactive --mode local \
  --auth-choice custom-api-key \
  --custom-base-url http://127.0.0.1:8080/v1 \
  --custom-model-id "archgod94/tellcue-models:Q4_K_M" \
  --custom-provider-id llama-cpp \
  --custom-compatibility openai \
  --custom-text-input \
  --accept-risk \
  --skip-health

Run OpenClaw

openclaw agent --local --agent main --message "Hello from Hugging Face"

Tellcue models

On-device models downloaded by the Tellcue video editor when a feature is first used.

rife-v4.17-lite.onnx

Frame interpolation for smooth slow motion.

Network and weights: RIFE v4.17 lite from Practical-RIFE (hzwer, MIT). Weights file flownet_v4.17.lite.pkl from the vs-rife model release (HolyWu, MIT).
Converted to ONNX (opset 17, fp32) with tool/export_rife.py in the Tellcue repository. Changes from the original: the flow grid is built inside the graph, so any size works; the flow scale is an input; the warps of blocks 1 to 3 run at the block size instead of full size, which is about twice as fast with the same result (within 0.07 dB on real frames).
Inputs: img0, img1 float32 [1, 3, H, W] RGB 0..1 (H and W multiples of 32 / scale), timestep float32 [1] in 0..1, scale float32 [1] (1, 0.5 or 0.25). Output: out float32 [1, 3, H, W].
SHA-256: c0887c5a56dba4f18a766f2c826ae889d92edb719b40e0732817da1241125889

deepfilternet3.onnx

Speech noise reduction (48 kHz).

Network and weights: DeepFilterNet3 by Hendrik Schroeter, models/DeepFilterNet3_onnx.tar.gz (MIT or Apache-2.0).
The three exported graphs (enc, erb_dec, df_dec) joined into one with tool/export_dfn.py in the Tellcue repository; numbers unchanged.
Inputs: feat_erb float32 [1, 1, S, 32], feat_spec float32 [1, 2, S, 96] (libDF features, 2 frames look-ahead applied). Outputs: m [1, 1, S, 32] ERB gains, coefs [1, S, 96, 10] deep filter taps, lsnr [1, S, 1].
SHA-256: 9223f1332c520e942b54a65a5e6e728ae56db7624b9da48c11f4585dc1afddce

migan-pipeline-v2.onnx

Object removal (inpainting) for photos.

MI-GAN (Picsart AI Research, MIT), the authors' pre-built ONNX pipeline migan_pipeline_v2.onnx as published at andraniksargsyan/migan, copied unchanged.
Inputs: image uint8 [1, 3, H, W] RGB, mask uint8 [1, 1, H, W] (0 = remove, 255 = keep), any size. Output: result uint8 [1, 3, H, W]; only the masked area changes.
SHA-256: 6f1f3530a1a2324b19752018ce756088b07973cda8d7d890034ace5c8a48c40b

realesr-general-x4v3.onnx

Photo upscaling (x4) with light denoising.

Real-ESRGAN by Xintao Wang et al. (BSD-3-Clause), weights realesr-general-x4v3.pth from the v0.2.5.0 release (SRVGGNetCompact, 64 features, 32 convolutions).
Converted to ONNX (opset 17, fp32, any size) with tool/export_esrgan.py in the Tellcue repository; output matches PyTorch within 6e-6.
Input: input float32 [1, 3, H, W] RGB 0..1. Output: output float32 [1, 3, 4H, 4W].
SHA-256: e859a54006946df52e810ea79c6ba71ef522ed1b6f8b04ae1646d66edddce293

mdx-kim-vocal-2.onnx

Vocal separation (isolate voice / remove vocals), 44.1 kHz stereo.

MDX-Net model Kim_Vocal_2.onnx trained by KimberleyJSN for Ultimate Vocal Remover (UVR, MIT), as distributed with UVR's model downloads; copied unchanged. The weights carry no licence of their own; UVR states MIT.
STFT settings (UVR model data): n_fft 7680, hop 1024, dim_f 3072, dim_t 256, compensate 1.009.
Input: input float32 [B, 4, 3072, 256] (left re, left im, right re, right im). Output: output same shape, the vocal spectrogram.
SHA-256: ce74ef3b6a6024ce44211a07be9cf8bc6d87728cc852a68ab34eb8e58cde9c8b

Kokoro text to speech (kokoro-*)

American and British English speech, 24 kHz. The front end (text to phonemes) runs in the app's own code from misaki's data; espeak-ng is not used.

kokoro-v1.0-int8.onnx: Kokoro-82M v1.0 by hexgrad (Apache-2.0), the int8 graph onnx/model_quantized.onnx of onnx-community/Kokoro-82M-v1.0-ONNX with one change (tool/build_kokoro.py): the token ids come in as float32 tokens [1, n] and are cast to int64 inside. Inputs: tokens, style float32 [1, 256], speed float32 [1]. Output: waveform float32 [1, samples]. Same samples as the original (max difference 0). SHA-256 90d4204097076bb335c662280fd1fb0d66811518f31d988e517a370c51ee98e3
kokoro-voices-v1.0.bin: 15 Kokoro voice packs (af_heart, af_bella, af_nicole, af_aoede, af_kore, af_sarah, am_michael, am_fenrir, am_puck, am_echo, bf_emma, bf_isabella, bm_george, bm_fable, bm_lewis), each float32 [510, 256], one after another. SHA-256 96919b8dd79b7c81d75142f3eff1c71bdc9856593100243131cf57e2f0366690
kokoro-lexicon-us.json.gz, kokoro-lexicon-gb.json.gz: the gold and silver pronunciation dictionaries of misaki 0.9.4 (hexgrad, Apache-2.0), unchanged, as {"gold": ..., "silver": ...}. SHA-256 19111eb337155965d209a792f830304455306c9b4f580470d6b412dd128726e3 / 0adb39c6ee1942702f5e0b5fd16162597c513dc9d9ebd9b20d991cd62e16b49d
kokoro-g2p-us.onnx, kokoro-g2p-gb.onnx: misaki's neural reader for unknown words, PeterReid/graphemes_to_phonemes_en_us / _gb (Apache-2.0, one-layer BART), exported with tool/export_g2p.py (opset 17, no cache). Inputs graphemes float32 [1, n] (1, letter ids, 2) and phonemes float32 [1, m] (decoder ids so far, from 1); output logits [1, m, 63]; greedy decoding gives the same result as transformers generate. SHA-256 e65f85732f5b78c804ca596015444066b81d4ca0f08625566bb4044982ce8fbb / be73a71fa46ddabe863bf4178ca543dba7edc4d27002b95b355a070ba751bbea

Ask Tellcue planner (qwen3-*.gguf)

Small language models that turn a typed request into an edit plan (JSON, constrained by a grammar) when the phone has no system language model. Run with llama.cpp b11430 inside the app.

qwen3-0.6b-q4_k_m.gguf: Qwen3-0.6B by the Qwen team, Alibaba Cloud (Apache-2.0), Q4_K_M quantisation from unsloth/Qwen3-0.6B-GGUF (Qwen3-0.6B-Q4_K_M.gguf), copied unchanged. Used on phones with less memory. SHA-256 ac2d97712095a558e31573f62f466a3f9d93990898b0ec79d7c974c1780d524a
qwen3-1.7b-q4_k_m.gguf: Qwen3-1.7B (Apache-2.0), Q4_K_M from unsloth/Qwen3-1.7B-GGUF (Qwen3-1.7B-Q4_K_M.gguf), copied unchanged. Used on phones with about 8 GB of memory or more. SHA-256 b139949c5bd74937ad8ed8c8cf3d9ffb1e99c866c823204dc42c0d91fa181897
Prompt format: ChatML with an empty <think> block (thinking off).

Magic Edit scene finder (yamnet.onnx, mobilenetv3-large-imagenet.onnx)

Sound and picture labels that tell Magic Edit what happens in a video without speech (music, sport, pets, places).

yamnet.onnx: YAMNet by Google (Apache-2.0), weights yamnet.h5 from storage.googleapis.com/audioset, rebuilt in PyTorch with its log-mel front end (the STFT as a fixed convolution) and converted to ONNX (opset 17, fp32) with tool/export_yamnet.py in the Tellcue repository; numbers unchanged. Input waveform float32 [B, 15600] (0.975 s of 16 kHz mono, -1..1). Output scores float32 [B, 521] (sigmoid, AudioSet classes in the order of yamnet_class_map.csv). SHA-256 f3941a025b9b81703bee7946fc7d8ba052d4ece1762c9df96ada50d3dece6ca2
mobilenetv3-large-imagenet.onnx: MobileNetV3-Large, torchvision IMAGENET1K_V2 weights (BSD-3-Clause), ImageNet normalisation and softmax inside, converted with tool/export_mobilenet.py. Input image float32 [B, 3, 224, 224] RGB 0..1. Output probs float32 [B, 1000] (ImageNet-1k classes in torchvision order). SHA-256 adbf157a1e4bdb83c605865c673ffb71e595083ce2d742cac791e4a63a154ae8

See LICENSE for the original licences.

Downloads last month: 57

GGUF

Model size

0.6B params

Architecture

qwen3

Hardware compatibility

4-bit