cstr/kartoffel-orpheus-3b-german-synthetic-GGUF

Kartoffel-Orpheus 3B (German, synthetic + emotions) — GGUF (ggml-quantised)

GGUF / ggml conversion of SebastianBodza/Kartoffel_Orpheus-3B_german_synthetic-v0.1 for use with CrispStrobe/CrispASR.

A German fine-tune of canopylabs/orpheus-3b-0.1-ft, trained on synthetic German speech with explicit emotion and outburst control. Drop-in checkpoint swap on the same Orpheus runtime — same Llama-3.2-3B-Instruct talker arch, same SNAC 24 kHz codec, same <custom_token_N> super-frame protocol. The natural-data sibling lives at cstr/kartoffel-orpheus-3b-german-natural-GGUF.

Speakers (4)

Name	Voice
Martin	Male
Luca	Male
Anne	Female
Emma	Female

Emotions + outbursts

The synthetic variant uses an extended prompt syntax:

{Speaker} - {Emotion}: {German text}

| Emotions | Neutral, Happy, Sad, Excited, Surprised, Humorous, Angry, Calm, Disgust, Fear, Proud, Romantic | | Outbursts | haha, ughh, wow, wuhuuu, ohhh (in-text, surrounded by spaces) |

Example: "Martin - Sad: Oh, ich bin so traurig." → mournful Martin voice. Example: "Anne - Happy: wow das ist ja großartig." → cheerful Anne with the wow outburst.

Pair this with the SNAC codec at cstr/snac-24khz-GGUF.

Files

File	Quant	Size	Notes
kartoffel-orpheus-de-synthetic-f16.gguf	F16	~6.2 GB	Reference quality
kartoffel-orpheus-de-synthetic-q8_0.gguf	Q8_0	~3.4 GB	Recommended
kartoffel-orpheus-de-synthetic-q4_k.gguf	Q4_K	~1.8 GB	Smallest

Quick start

# 1. Build CrispASR
git clone https://github.com/CrispStrobe/CrispASR
cd CrispASR
cmake -B build -DCMAKE_BUILD_TYPE=Release
cmake --build build -j --target crispasr-lib

# 2. Pull the talker + the SNAC codec
huggingface-cli download cstr/kartoffel-orpheus-3b-german-synthetic-GGUF \
    kartoffel-orpheus-de-synthetic-q8_0.gguf --local-dir .
huggingface-cli download cstr/snac-24khz-GGUF snac-24khz.gguf --local-dir .

# 3. Synthesise — German prompt with emotion control
./build/bin/crispasr --backend kartoffel-orpheus \
    -m kartoffel-orpheus-de-synthetic-q8_0.gguf \
    --codec-model snac-24khz.gguf \
    --voice Martin \
    --temperature 0.6 \
    --tts "Martin - Sad: Oh, ich bin so traurig." \
    --tts-output martin_sad.wav

For auto-download simply pass -m auto:

./build/bin/crispasr --backend kartoffel-orpheus-de-synthetic -m auto \
    --voice Anne --temperature 0.6 \
    --tts "Anne - Happy: Hallo, wie geht es dir heute?" \
    --tts-output anne_happy.wav

Architecture

Identical to Orpheus 3B — see cstr/orpheus-3b-0.1-ft-GGUF for the full architecture writeup. The CrispASR orpheus runtime is checkpoint-agnostic; this GGUF is loaded by the same orpheus_init_from_file path with no source-code changes. The --voice flag here is just the speaker name string in the prompt; emotion and outburst control happens via the prompt text itself.

Conversion

python models/convert-orpheus-to-gguf.py \
    --input SebastianBodza/Kartoffel_Orpheus-3B_german_synthetic-v0.1 \
    --output kartoffel-orpheus-de-synthetic-f16.gguf \
    --speakers Martin,Luca,Anne,Emma \
    --variant fixed_speaker

build/bin/crispasr-quantize kartoffel-orpheus-de-synthetic-f16.gguf \
    kartoffel-orpheus-de-synthetic-q8_0.gguf q8_0
build/bin/crispasr-quantize kartoffel-orpheus-de-synthetic-f16.gguf \
    kartoffel-orpheus-de-synthetic-q4_k.gguf q4_k

Attribution

• Talker base: SebastianBodza/Kartoffel_Orpheus-3B_german_synthetic-v0.1.
• Upstream Orpheus base: canopylabs/orpheus-3b-0.1-ft.
• Llama base: meta-llama/Llama-3.2-3B-Instruct — Llama-3.2 community license.
• SNAC codec: hubertsiuzdak/snac_24khz (MIT) — see cstr/snac-24khz-GGUF.
• GGUF conversion + ggml runtime: CrispStrobe/CrispASR — see src/orpheus.cpp, src/orpheus_snac.cpp, models/convert-orpheus-to-gguf.py.

License

Llama-3.2 community license (inherited from the talker base). Includes the Acceptable Use Policy and the "Built with Llama" attribution requirement.

The SNAC codec is MIT and ships separately under cstr/snac-24khz-GGUF.